Links to Exploration step
Le document en format XML
<record><TEI><teiHeader><fileDesc><titleStmt><title xml:lang="en">Threshold Models for Genome-Enabled Prediction of Ordinal Categorical Traits in Plant Breeding</title><author><name sortKey="Montesinos L Pez, Osval A" sort="Montesinos L Pez, Osval A" uniqKey="Montesinos L Pez O" first="Osval A." last="Montesinos-L Pez">Osval A. Montesinos-L Pez</name><affiliation><nlm:aff id="aff1">Facultad de Telemática, Universidad de Colima, Avenida Universidad 333, Colima, México</nlm:aff></affiliation></author><author><name sortKey="Montesinos L Pez, Abelardo" sort="Montesinos L Pez, Abelardo" uniqKey="Montesinos L Pez A" first="Abelardo" last="Montesinos-L Pez">Abelardo Montesinos-L Pez</name><affiliation><nlm:aff id="aff2">Departamento de Estadística, Centro de Investigación en Matemáticas (CIMAT), Guanajuato, Guanajuato, México</nlm:aff></affiliation></author><author><name sortKey="Perez Rodriguez, Paulino" sort="Perez Rodriguez, Paulino" uniqKey="Perez Rodriguez P" first="Paulino" last="Pérez-Rodríguez">Paulino Pérez-Rodríguez</name><affiliation><nlm:aff id="aff3">Colegio de Postgraduados, Edo. de México, México</nlm:aff></affiliation></author><author><name sortKey="De Los Campos, Gustavo" sort="De Los Campos, Gustavo" uniqKey="De Los Campos G" first="Gustavo" last="De Los Campos">Gustavo De Los Campos</name><affiliation><nlm:aff id="aff4">Department of Biostatistics, University of Alabama at Birmingham, 443, Birmingham, Alabama 35294</nlm:aff></affiliation></author><author><name sortKey="Eskridge, Kent" sort="Eskridge, Kent" uniqKey="Eskridge K" first="Kent" last="Eskridge">Kent Eskridge</name><affiliation><nlm:aff id="aff5">Department of Statistics, University of Nebraska, Lincoln, Nebraska</nlm:aff></affiliation></author><author><name sortKey="Crossa, Jose" sort="Crossa, Jose" uniqKey="Crossa J" first="José" last="Crossa">José Crossa</name><affiliation><nlm:aff id="aff6">Biometrics and Statistics Unit, International Maize and Wheat Improvement Center (CIMMYT), México, D.F., México</nlm:aff></affiliation></author></titleStmt><publicationStmt><idno type="wicri:source">PMC</idno><idno type="pmid">25538102</idno><idno type="pmc">4321037</idno><idno type="url">http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4321037</idno><idno type="RBID">PMC:4321037</idno><idno type="doi">10.1534/g3.114.016188</idno><date when="2014">2014</date><idno type="wicri:Area/Pmc/Corpus">000452</idno><idno type="wicri:explorRef" wicri:stream="Pmc" wicri:step="Corpus" wicri:corpus="PMC">000452</idno></publicationStmt><sourceDesc><biblStruct><analytic><title xml:lang="en" level="a" type="main">Threshold Models for Genome-Enabled Prediction of Ordinal Categorical Traits in Plant Breeding</title><author><name sortKey="Montesinos L Pez, Osval A" sort="Montesinos L Pez, Osval A" uniqKey="Montesinos L Pez O" first="Osval A." last="Montesinos-L Pez">Osval A. Montesinos-L Pez</name><affiliation><nlm:aff id="aff1">Facultad de Telemática, Universidad de Colima, Avenida Universidad 333, Colima, México</nlm:aff></affiliation></author><author><name sortKey="Montesinos L Pez, Abelardo" sort="Montesinos L Pez, Abelardo" uniqKey="Montesinos L Pez A" first="Abelardo" last="Montesinos-L Pez">Abelardo Montesinos-L Pez</name><affiliation><nlm:aff id="aff2">Departamento de Estadística, Centro de Investigación en Matemáticas (CIMAT), Guanajuato, Guanajuato, México</nlm:aff></affiliation></author><author><name sortKey="Perez Rodriguez, Paulino" sort="Perez Rodriguez, Paulino" uniqKey="Perez Rodriguez P" first="Paulino" last="Pérez-Rodríguez">Paulino Pérez-Rodríguez</name><affiliation><nlm:aff id="aff3">Colegio de Postgraduados, Edo. de México, México</nlm:aff></affiliation></author><author><name sortKey="De Los Campos, Gustavo" sort="De Los Campos, Gustavo" uniqKey="De Los Campos G" first="Gustavo" last="De Los Campos">Gustavo De Los Campos</name><affiliation><nlm:aff id="aff4">Department of Biostatistics, University of Alabama at Birmingham, 443, Birmingham, Alabama 35294</nlm:aff></affiliation></author><author><name sortKey="Eskridge, Kent" sort="Eskridge, Kent" uniqKey="Eskridge K" first="Kent" last="Eskridge">Kent Eskridge</name><affiliation><nlm:aff id="aff5">Department of Statistics, University of Nebraska, Lincoln, Nebraska</nlm:aff></affiliation></author><author><name sortKey="Crossa, Jose" sort="Crossa, Jose" uniqKey="Crossa J" first="José" last="Crossa">José Crossa</name><affiliation><nlm:aff id="aff6">Biometrics and Statistics Unit, International Maize and Wheat Improvement Center (CIMMYT), México, D.F., México</nlm:aff></affiliation></author></analytic><series><title level="j">G3: Genes|Genomes|Genetics</title><idno type="eISSN">2160-1836</idno><imprint><date when="2014">2014</date></imprint></series></biblStruct></sourceDesc></fileDesc><profileDesc><textClass></textClass></profileDesc></teiHeader><front><div type="abstract" xml:lang="en"><p>Categorical scores for disease susceptibility or resistance often are recorded in plant breeding. The aim of this study was to introduce genomic models for analyzing ordinal characters and to assess the predictive ability of genomic predictions for ordered categorical phenotypes using a threshold model counterpart of the Genomic Best Linear Unbiased Predictor (<italic>i.e.</italic>, TGBLUP). The threshold model was used to relate a hypothetical underlying scale to the outward categorical response. We present an empirical application where a total of nine models, five without interaction and four with genomic × environment interaction (G×E) and genomic additive × additive × environment interaction (G×G×E), were used. We assessed the proposed models using data consisting of 278 maize lines genotyped with 46,347 single-nucleotide polymorphisms and evaluated for disease resistance [with ordinal scores from 1 (no disease) to 5 (complete infection)] in three environments (Colombia, Zimbabwe, and Mexico). Models with G×E captured a sizeable proportion of the total variability, which indicates the importance of introducing interaction to improve prediction accuracy. Relative to models based on main effects only, the models that included G×E achieved 9–14% gains in prediction accuracy; adding additive × additive interactions did not increase prediction accuracy consistently across locations.</p></div></front><back><div1 type="bibliography"><listBibl><biblStruct><analytic><author><name sortKey="Atkinson, L" uniqKey="Atkinson L">L. Atkinson</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Brier, G W" uniqKey="Brier G">G. W. Brier</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Burgue O, J" uniqKey="Burgue O J">J. Burgueño</name></author><author><name sortKey="De Los Campos, G" uniqKey="De Los Campos G">G. de los Campos</name></author><author><name sortKey="Weigel, K" uniqKey="Weigel K">K. Weigel</name></author><author><name sortKey="Crossa, J" uniqKey="Crossa J">J. Crossa</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Crossa, J" uniqKey="Crossa J">J. Crossa</name></author><author><name sortKey="De Los Campos, G" uniqKey="De Los Campos G">G. de los Campos</name></author><author><name sortKey="Perez Rodriguez, P" uniqKey="Perez Rodriguez P">P. Pérez-Rodríguez</name></author><author><name sortKey="Gianola, D" uniqKey="Gianola D">D. Gianola</name></author><author><name sortKey="Burgue O, J" uniqKey="Burgue O J">J. Burgueño</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Crossa, J" uniqKey="Crossa J">J. Crossa</name></author><author><name sortKey="Perez Rodriguez, P" uniqKey="Perez Rodriguez P">P. Pérez-Rodríguez</name></author><author><name sortKey="De Los Campos, G" uniqKey="De Los Campos G">G. de los Campos</name></author><author><name sortKey="Mahuku, G" uniqKey="Mahuku G">G. Mahuku</name></author><author><name sortKey="Dreisigacker, S" uniqKey="Dreisigacker S">S. Dreisigacker</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Crossa, J" uniqKey="Crossa J">J. Crossa</name></author><author><name sortKey="Beyene, Y" uniqKey="Beyene Y">Y. Beyene</name></author><author><name sortKey="Kassa, S" uniqKey="Kassa S">S. Kassa</name></author><author><name sortKey="Perez Rodriguez, P" uniqKey="Perez Rodriguez P">P. Pérez-Rodríguez</name></author><author><name sortKey="Hickey, J M" uniqKey="Hickey J">J. M. Hickey</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Crossa, J" uniqKey="Crossa J">J. Crossa</name></author><author><name sortKey="Perez Rodriguez, P" uniqKey="Perez Rodriguez P">P. Pérez-Rodríguez</name></author><author><name sortKey="Hickey, J" uniqKey="Hickey J">J. Hickey</name></author><author><name sortKey="Burgue O, J" uniqKey="Burgue O J">J. Burgueño</name></author><author><name sortKey="Ornella, L" uniqKey="Ornella L">L. Ornella</name></author></analytic></biblStruct><biblStruct></biblStruct><biblStruct><analytic><author><name sortKey="De Los Campos, G" uniqKey="De Los Campos G">G. de los Campos</name></author><author><name sortKey="Naya, H" uniqKey="Naya H">H. Naya</name></author><author><name sortKey="Gianola, D" uniqKey="Gianola D">D. Gianola</name></author><author><name sortKey="Crossa, J" uniqKey="Crossa J">J. Crossa</name></author><author><name sortKey="Legarra, A" uniqKey="Legarra A">A. Legarra</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="De Los Campos, G" uniqKey="De Los Campos G">G. de los Campos</name></author><author><name sortKey="Gianola, D" uniqKey="Gianola D">D. Gianola</name></author><author><name sortKey="Rosa, G J M" uniqKey="Rosa G">G. J. M. Rosa</name></author><author><name sortKey="Weigel, K" uniqKey="Weigel K">K. Weigel</name></author><author><name sortKey="Crossa, J" uniqKey="Crossa J">J. Crossa</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Garthwaite, P H" uniqKey="Garthwaite P">P. H. Garthwaite</name></author><author><name sortKey="Kadane, J B" uniqKey="Kadane J">J. B. Kadane</name></author><author><name sortKey="O Agan, A" uniqKey="O Agan A">A. O’Hagan</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Gianola, D" uniqKey="Gianola D">D. Gianola</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Gianola, D" uniqKey="Gianola D">D. Gianola</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Gianola, D" uniqKey="Gianola D">D. Gianola</name></author><author><name sortKey="Foulley, J L" uniqKey="Foulley J">J. L. Foulley</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Gonzalez Camacho, J M" uniqKey="Gonzalez Camacho J">J. M. González-Camacho</name></author><author><name sortKey="De Los Campos, G" uniqKey="De Los Campos G">G. de los Campos</name></author><author><name sortKey="Perez Rodriguez, P" uniqKey="Perez Rodriguez P">P. Pérez-Rodríguez</name></author><author><name sortKey="Gianola, D" uniqKey="Gianola D">D. Gianola</name></author><author><name sortKey="Cairns, J E" uniqKey="Cairns J">J. E. Cairns</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Gonzalez Recio, O" uniqKey="Gonzalez Recio O">O. González-Recio</name></author><author><name sortKey="Forni, S" uniqKey="Forni S">S. Forni</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Heffner, E L" uniqKey="Heffner E">E. L. Heffner</name></author><author><name sortKey="Sorrells, M E" uniqKey="Sorrells M">M. E. Sorrells</name></author><author><name sortKey="Jannink, J L" uniqKey="Jannink J">J.-L. Jannink</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Heffner, E L" uniqKey="Heffner E">E. L. Heffner</name></author><author><name sortKey="Lorenz, A J" uniqKey="Lorenz A">A. J. Lorenz</name></author><author><name sortKey="Jannink, J L" uniqKey="Jannink J">J.-L. Jannink</name></author><author><name sortKey="Sorrells, M E" uniqKey="Sorrells M">M. E. Sorrells</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Heslot, N" uniqKey="Heslot N">N. Heslot</name></author><author><name sortKey="Yang, H P" uniqKey="Yang H">H.-P. Yang</name></author><author><name sortKey="Sorrells, M E" uniqKey="Sorrells M">M. E. Sorrells</name></author><author><name sortKey="Jannink, J L" uniqKey="Jannink J">J.-L. Jannink</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Jarquin, D" uniqKey="Jarquin D">D. Jarquín</name></author><author><name sortKey="Crossa, J" uniqKey="Crossa J">J. Crossa</name></author><author><name sortKey="Lacaze, X" uniqKey="Lacaze X">X. Lacaze</name></author><author><name sortKey="Cheyron, P" uniqKey="Cheyron P">P. Cheyron</name></author><author><name sortKey="Daucourt, J" uniqKey="Daucourt J">J. Daucourt</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Kizilkaya, K" uniqKey="Kizilkaya K">K. Kizilkaya</name></author><author><name sortKey="Tait, R G" uniqKey="Tait R">R. G. Tait</name></author><author><name sortKey="Garrick, D J" uniqKey="Garrick D">D. J. Garrick</name></author><author><name sortKey="Fernando, R L" uniqKey="Fernando R">R. L. Fernando</name></author><author><name sortKey="Reecy, J M" uniqKey="Reecy J">J. M. Reecy</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Kneib, T" uniqKey="Kneib T">T. Kneib</name></author><author><name sortKey="Baumgartner, B" uniqKey="Baumgartner B">B. Baumgartner</name></author><author><name sortKey="Steiner, W J" uniqKey="Steiner W">W. J. Steiner</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Littell, R C" uniqKey="Littell R">R. C. Littell</name></author><author><name sortKey="Stroup, W" uniqKey="Stroup W">W. Stroup</name></author><author><name sortKey="Freund, R J" uniqKey="Freund R">R. J. Freund</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Meuwissen, T H E" uniqKey="Meuwissen T">T. H. E. Meuwissen</name></author><author><name sortKey="Hayes, B J" uniqKey="Hayes B">B. J. Hayes</name></author><author><name sortKey="Goddard, M E" uniqKey="Goddard M">M. E. Goddard</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Perez Rodriguez, P" uniqKey="Perez Rodriguez P">P. Pérez-Rodríguez</name></author><author><name sortKey="De Los Campos, G" uniqKey="De Los Campos G">G. de los Campos</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Perez Rodriguez, P" uniqKey="Perez Rodriguez P">P. Pérez-Rodríguez</name></author><author><name sortKey="De Los Campos, G" uniqKey="De Los Campos G">G. de los Campos</name></author><author><name sortKey="Crossa, J" uniqKey="Crossa J">J. Crossa</name></author><author><name sortKey="Gianola, D" uniqKey="Gianola D">D. Gianola</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Perez Rodriguez, P" uniqKey="Perez Rodriguez P">P. Pérez-Rodríguez</name></author><author><name sortKey="Gianola, D" uniqKey="Gianola D">D. Gianola</name></author><author><name sortKey="Gonzalez Camacho, J M" uniqKey="Gonzalez Camacho J">J. M. González-Camacho</name></author><author><name sortKey="Crossa, J" uniqKey="Crossa J">J. Crossa</name></author><author><name sortKey="Manes, Y" uniqKey="Manes Y">Y. Manes</name></author><author><name sortKey="Dreisigacker, S" uniqKey="Dreisigacker S">S. Dreisigacker,</name></author></analytic></biblStruct><biblStruct></biblStruct><biblStruct><analytic><author><name sortKey="Sorensen, D A" uniqKey="Sorensen D">D. A. Sorensen</name></author><author><name sortKey="Andersen, S" uniqKey="Andersen S">S. Andersen</name></author><author><name sortKey="Gianola, D" uniqKey="Gianola D">D. Gianola</name></author><author><name sortKey="Korsgaard, I" uniqKey="Korsgaard I">I. Korsgaard</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Stroup, W" uniqKey="Stroup W">W. Stroup</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Villanueva, B" uniqKey="Villanueva B">B. Villanueva</name></author><author><name sortKey="Fernandez, J" uniqKey="Fernandez J">J. Fernandez</name></author><author><name sortKey="Garcia Cortes, L A" uniqKey="Garcia Cortes L">L. A. Garcia-Cortes</name></author><author><name sortKey="Varona, L" uniqKey="Varona L">L. Varona</name></author><author><name sortKey="Daetwyler, H D" uniqKey="Daetwyler H">H. D. Daetwyler</name></author></analytic></biblStruct><biblStruct><analytic><author><name sortKey="Wang, C L" uniqKey="Wang C">C. L. Wang</name></author><author><name sortKey="Ding, X D" uniqKey="Ding X">X. D. Ding</name></author><author><name sortKey="Wang, J Y" uniqKey="Wang J">J. Y. Wang</name></author><author><name sortKey="Liu, J F" uniqKey="Liu J">J. F. Liu</name></author><author><name sortKey="Fu, W X" uniqKey="Fu W">W. X. Fu</name></author></analytic></biblStruct></listBibl></div1></back></TEI><pmc article-type="research-article"><pmc-dir>properties open_access</pmc-dir>
<front><journal-meta><journal-id journal-id-type="nlm-ta">G3 (Bethesda)</journal-id><journal-id journal-id-type="iso-abbrev">Genetics</journal-id><journal-id journal-id-type="hwp">G3: Genes, Genomes, Genetics</journal-id><journal-id journal-id-type="pmc">G3: Genes, Genomes, Genetics</journal-id><journal-id journal-id-type="publisher-id">G3: Genes, Genomes, Genetics</journal-id><journal-title-group><journal-title>G3: Genes|Genomes|Genetics</journal-title></journal-title-group><issn pub-type="epub">2160-1836</issn><publisher><publisher-name>Genetics Society of America</publisher-name></publisher></journal-meta><article-meta><article-id pub-id-type="pmid">25538102</article-id><article-id pub-id-type="pmc">4321037</article-id><article-id pub-id-type="publisher-id">GGG_016188</article-id><article-id pub-id-type="doi">10.1534/g3.114.016188</article-id><article-categories><subj-group subj-group-type="heading"><subject>Genomic Selection</subject></subj-group></article-categories><title-group><article-title>Threshold Models for Genome-Enabled Prediction of Ordinal Categorical Traits in Plant Breeding</article-title></title-group><contrib-group><contrib contrib-type="author"><name><surname>Montesinos-López</surname><given-names>Osval A.</given-names></name><xref ref-type="aff" rid="aff1">*</xref></contrib><contrib contrib-type="author"><name><surname>Montesinos-López</surname><given-names>Abelardo</given-names></name><xref ref-type="aff" rid="aff2"><sup>†</sup></xref></contrib><contrib contrib-type="author"><name><surname>Pérez-Rodríguez</surname><given-names>Paulino</given-names></name><xref ref-type="aff" rid="aff3"><sup>‡</sup></xref></contrib><contrib contrib-type="author"><name><surname>de los Campos</surname><given-names>Gustavo</given-names></name><xref ref-type="aff" rid="aff4"><sup>§</sup></xref></contrib><contrib contrib-type="author"><name><surname>Eskridge</surname><given-names>Kent</given-names></name><xref ref-type="aff" rid="aff5">**</xref></contrib><contrib contrib-type="author"><name><surname>Crossa</surname><given-names>José</given-names></name><xref ref-type="aff" rid="aff6"><sup>††</sup></xref><xref ref-type="corresp" rid="cor1"><sup>1</sup></xref></contrib><aff id="aff1"><label>*</label>Facultad de Telemática, Universidad de Colima, Avenida Universidad 333, Colima, México</aff><aff id="aff2"><label>†</label>Departamento de Estadística, Centro de Investigación en Matemáticas (CIMAT), Guanajuato, Guanajuato, México</aff><aff id="aff3"><label>‡</label>Colegio de Postgraduados, Edo. de México, México</aff><aff id="aff4"><label>§</label>Department of Biostatistics, University of Alabama at Birmingham, 443, Birmingham, Alabama 35294</aff><aff id="aff5"><label>**</label>Department of Statistics, University of Nebraska, Lincoln, Nebraska</aff><aff id="aff6"><label>††</label>Biometrics and Statistics Unit, International Maize and Wheat Improvement Center (CIMMYT), México, D.F., México</aff></contrib-group><author-notes><corresp id="cor1"><label>1</label>Corresponding author: Biometrics and Statistics Unit, International Maize and Wheat Improvement Center (CIMMYT), Apdo. Postal 6-641, México, D.F., México. E-mail: <email>j.crossa@cgiar.org</email></corresp></author-notes><pub-date pub-type="epub"><day>23</day><month>12</month><year>2014</year></pub-date><pub-date pub-type="collection"><month>2</month><year>2015</year></pub-date><volume>5</volume><issue>2</issue><fpage>291</fpage><lpage>300</lpage><history><date date-type="received"><day>31</day><month>10</month><year>2014</year></date><date date-type="accepted"><day>19</day><month>12</month><year>2014</year></date></history><permissions><copyright-statement>Copyright © 2015 Montesinos-López <italic>et al</italic>.</copyright-statement><copyright-year>2015</copyright-year><license license-type="open-access" xlink:href="http://creativecommons.org/licenses/by/3.0/"><license-p>This is an open-access article distributed under the terms of the Creative Commons Attribution Unported License (<ext-link ext-link-type="uri" xlink:href="http://creativecommons.org/licenses/by/3.0/">http://creativecommons.org/licenses/by/3.0/</ext-link>), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.</license-p></license></permissions><self-uri xlink:title="pdf" xlink:type="simple" xlink:href="291.pdf"></self-uri><abstract><p>Categorical scores for disease susceptibility or resistance often are recorded in plant breeding. The aim of this study was to introduce genomic models for analyzing ordinal characters and to assess the predictive ability of genomic predictions for ordered categorical phenotypes using a threshold model counterpart of the Genomic Best Linear Unbiased Predictor (<italic>i.e.</italic>, TGBLUP). The threshold model was used to relate a hypothetical underlying scale to the outward categorical response. We present an empirical application where a total of nine models, five without interaction and four with genomic × environment interaction (G×E) and genomic additive × additive × environment interaction (G×G×E), were used. We assessed the proposed models using data consisting of 278 maize lines genotyped with 46,347 single-nucleotide polymorphisms and evaluated for disease resistance [with ordinal scores from 1 (no disease) to 5 (complete infection)] in three environments (Colombia, Zimbabwe, and Mexico). Models with G×E captured a sizeable proportion of the total variability, which indicates the importance of introducing interaction to improve prediction accuracy. Relative to models based on main effects only, the models that included G×E achieved 9–14% gains in prediction accuracy; adding additive × additive interactions did not increase prediction accuracy consistently across locations.</p></abstract><kwd-group><kwd>prediction accuracy</kwd><kwd>threshold model</kwd><kwd>disease resistance</kwd><kwd>GBLUP</kwd><kwd>genotype × environment interaction</kwd><kwd>GenPred</kwd><kwd>shared data resource</kwd></kwd-group><counts><page-count count="10"></page-count></counts><custom-meta-group><custom-meta><meta-name> DJS Export </meta-name><meta-value>v1</meta-value></custom-meta></custom-meta-group></article-meta></front><body><p>Since genomic selection (GS) initially was proposed by <xref rid="bib24" ref-type="bibr">Meuwissen <italic>et al.</italic> (2001)</xref>, a large number of plant breeding studies have assessed the prediction accuracy of GS in different economically important crops. These studies have used different marker platforms and marker densities, as well as different parametric and nonparametric statistical models (<italic>e.g.</italic>, <xref rid="bib9" ref-type="bibr">de los Campos <italic>et al.</italic> 2009</xref>, <xref rid="bib10" ref-type="bibr">2010</xref>; <xref rid="bib17" ref-type="bibr">Heffner <italic>et al.</italic> 2009</xref>, <xref rid="bib18" ref-type="bibr">2010</xref>; <xref rid="bib4" ref-type="bibr">Crossa <italic>et al.</italic> 2010</xref>, <xref rid="bib5" ref-type="bibr">2011</xref>, <xref rid="bib6" ref-type="bibr">2013a</xref>,<xref rid="bib7" ref-type="bibr">b</xref>; <xref rid="bib19" ref-type="bibr">Heslot <italic>et al.</italic> 2012</xref>; <xref rid="bib26" ref-type="bibr">Pérez-Rodríguez <italic>et al.</italic> 2010</xref>, <xref rid="bib27" ref-type="bibr">2012</xref>). Both simulation and empirical studies have shown that GS has greater prediction accuracy than standard pedigree-based prediction, and most of the benefits of GS arise from obtaining accurate predictions early in the breeding cycle. The success of genomic prediction is affected by the choice of model, the size of the training data, the heritability of the trait, the span of linkage disequilibrium, the marker density, and the strength of the genetic relationships between the training and validation populations, among other factors. A number of studies have assessed how these factors affect prediction accuracy for quantitative traits (<xref rid="bib21" ref-type="bibr">Kizilkaya <italic>et al.</italic> 2014</xref>); however, little has been done with categorical traits. As first pointed out by <xref rid="bib12" ref-type="bibr">Gianola (1980</xref>, <xref rid="bib13" ref-type="bibr">1982)</xref> and <xref rid="bib14" ref-type="bibr">Gianola and Foulley (1983)</xref>, these traits cannot be dealt with by the use of Gaussian models.</p><p>Several studies support the notion that when the number of categories is large and the data seem to follow an approximately normal distribution, failure to address the ordinality of the data is likely negligible (<xref rid="bib1" ref-type="bibr">Atkinson 1988</xref>). However, significant bias is observed when the number of categories is less than five and sample size is not large enough. In addition, analyzing the data as normally distributed has less power of detection of effects and often produces estimates of frequency categories outside the 0−1 range, which does not make sense. These problems often occur when the frequencies are small or very large. Also, even when treatment means obtained with a normal approximation can be interpreted as estimates of the proportions of each category, this is not the case for the variance of their estimates as the data are binary and polychotomous (<xref rid="bib30" ref-type="bibr">Stroup 2012</xref>). If the data are transformed, many of the aforementioned problems remain in a linear model analysis, given that the most commonly used transformations are intended to stabilize variance but fail to address the problem of skewness. Consequently, transformations often are ineffective and, in addition, express the data on scales that are unfamiliar to those who use the results of the analysis (<xref rid="bib23" ref-type="bibr">Littell <italic>et al.</italic> 2002</xref>).</p><p>Categorical traits are scored by assigning a data point to one of several mutually exclusive and exhaustive ordered categories. If these scores are treated as continuous variables, as in standard GS linear models, the following assumptions do not hold: (1) the relationship between genomic data and phenotypes is linear; (2) phenotypes follow a normal distribution; and (3) the variance is constant and not a function of the expected value. Therefore, standard GS linear models currently used in plant and animal breeding do not meet the assumptions required for categorical data.</p><p>Results of empirical and simulation studies indicate that generalized linear mixed models (GLMMs) provide more sensible results and have greater power to identify model effects as statistically significant (<xref rid="bib30" ref-type="bibr">Stroup 2012</xref>). In a linear model, the response variable (observation) equals the sum of explanatory variables plus a residual, and a probability assumption about the residual is made. In GLMM, the statistical model for a multinomial variable is expressed in a probability distribution form. It is important to point out that the cumulative probit model (also known as the threshold model) assumes that the process that gives rise to the observed categories is an underlying continuous variable with a standard normal distribution that in many applications uses the linear predictor with reversed signs for the fixed and random effects: <inline-formula><mml:math id="me1"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">η</mml:mi><mml:mi>c</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:msub><mml:mi>γ</mml:mi><mml:mi>c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mi mathvariant="bold-italic">Xβ</mml:mi><mml:mo>+</mml:mo><mml:mi mathvariant="bold-italic">Zb</mml:mi></mml:mrow></mml:math></inline-formula>.</p><p>Threshold models have been used in animal GS to relate a hypothetical underlying scale to the outward categorical response. For example, <xref rid="bib16" ref-type="bibr">González-Recio and Forni (2011)</xref> developed versions of BayesA, Bayesian Lasso, and two machine-learning methods for analyzing dichotomous traits and showed that the differences between methods are small, particularly with a large number of quantitative trait loci. On the other hand, <xref rid="bib31" ref-type="bibr">Villanueva <italic>et al.</italic> (2011)</xref> developed a version of the BayesB method for dichotomous traits and concluded that the threshold BayesB method improves GS accuracy when disease-resistant dichotomous phenotypes are encountered, compared with accuracies obtained with the linear model. The threshold model showed an increase in accuracy of up to 16%, as well as significant advantages when heritability and disease prevalence were low and individuals were genotyped but not measured (testing set). Both <xref rid="bib16" ref-type="bibr">González-Recio and Forni (2011)</xref> and <xref rid="bib31" ref-type="bibr">Villanueva <italic>et al.</italic> (2011)</xref> pointed out that the models they developed for dichotomous phenotypes can easily be extended to traits with more than two discrete categories.</p><p>Recently, <xref rid="bib32" ref-type="bibr">Wang <italic>et al.</italic> (2012)</xref> proposed threshold models using the GS framework and extended three Bayesian methods (BayesA, BayesB, and BayesC<inline-formula><mml:math id="me2"><mml:mi>π</mml:mi></mml:math></inline-formula>) to estimate genomic breeding values of animal threshold traits with more than two categories; they named these methods extended BayesTA, extended BayesTB, and extended BayesTC<inline-formula><mml:math id="me3"><mml:mi>π</mml:mi></mml:math></inline-formula>. They derived computing procedures for the three BayesT methods by using the Gibbs sampler algorithm. Through a simulation study, they found that the three BayesT methods generally performed better than the corresponding standard Bayesian methods, especially when only two phenotypic categories are present. They also found that BayesTB and BayesTC<inline-formula><mml:math id="me4"><mml:mi>π</mml:mi></mml:math></inline-formula> produced similar accuracies and that both performed better than BayesTA. Also, they addressed how heritability, number of quantitative trait loci, incidence, and number of phenotypic categories affect the performance of the three BayesT methods. Recently, <xref rid="bib21" ref-type="bibr">Kizilkaya <italic>et al.</italic> (2014)</xref> proposed performing genomic prediction of ordinal categorical phenotypes using BayesTC<inline-formula><mml:math id="me5"><mml:mi>π</mml:mi></mml:math></inline-formula> and compared it with BayesC<inline-formula><mml:math id="me6"><mml:mi>π</mml:mi></mml:math></inline-formula>; they found that the accuracies of BayesTC<inline-formula><mml:math id="me7"><mml:mi>π</mml:mi></mml:math></inline-formula> for ordinal categorical phenotypes were greater than those of BayesC<inline-formula><mml:math id="me8"><mml:mi>π</mml:mi></mml:math></inline-formula> and that there was a greater advantage in using BayesTC<inline-formula><mml:math id="me9"><mml:mi>π</mml:mi></mml:math></inline-formula> when the training population was small. These authors pointed out that a 2.25-fold increase in the training population size for ordinal categorical phenotypes analyzed as a linear model using BayesC<inline-formula><mml:math id="me10"><mml:mrow><mml:mi>π</mml:mi></mml:mrow></mml:math></inline-formula> was sufficient to achieve an accuracy equal to or greater than that for continuous phenotypes with a training population size of 1000 individuals.</p><p>In plant breeding, the most economically important trait is grain yield, which is greatly affected by environmental factors, as well as by genotype × environment interaction (G×E). The GS models first introduced by <xref rid="bib24" ref-type="bibr">Meuwissen <italic>et al.</italic> (2001)</xref> were further extended into multienvironment models that can handle large numbers of individuals genotyped with large numbers of markers and evaluated in multiple environments. <xref rid="bib3" ref-type="bibr">Burgueño <italic>et al.</italic> (2012)</xref> analyzed multienvironment data using a multivariate version of the genomic best linear unbiased predictor (GBLUP). <xref rid="bib20" ref-type="bibr">Jarquín <italic>et al.</italic> (2014)</xref> extended GBLUP to incorporate highly dimensional markers, environmental covariates, and their interactions into a class of random-effects models where main and interaction effects are modeled using Gaussian processes with covariance functions based on genetic and environmental similarity among entries. In their study, <xref rid="bib20" ref-type="bibr">Jarquín <italic>et al.</italic> (2014)</xref> used as a covariance function the structure induced by a reaction norm model, and applied the proposed approach to the grain yield trait in wheat lines evaluated over multiple years and locations.</p><p>However, in plant breeding, many economically important traits, such as the degree of resistance/susceptibility, are categorical. Despite this, to our knowledge, threshold models have not been considered in GS for plant breeding. Therefore, in this study, we introduced a threshold GS model that is an extension of the GBLUP of <xref rid="bib20" ref-type="bibr">Jarquín <italic>et al.</italic> (2014)</xref> which incorporates G×E and additive × additive × environment (G×G×E) interactions. The proposed model was evaluated using a data set consisting of 278 maize lines scored for resistance to gray leaf spot (GLS) measured using an ordered categorical scale [1 (no disease) to 5 (complete infection)] in three environments and genetic information consisting of 46,347 single-nucleotide polymorphisms (SNPs).</p><sec sec-type="materials|methods" id="s1"><title>Materials and Methods</title><sec id="s2"><title>Experimental data</title><sec id="s3"><title>Phenotypic data:</title><p>The trait analyzed in 278 maize lines was gray leaf spot (GLS), caused by the fungus <italic>Cercospora zeae-maydis</italic>, which was evaluated in three environments, Mexico, Zimbabwe, and Colombia. GLS is one of the most important foliar diseases of maize worldwide. The GLS trait was analyzed using an ordinal scale from 1 (no disease), 2 (low infection), 3 (moderate infection), 4 (high infection), to 5 (complete infection). These data are part of the data set previously analyzed by <xref rid="bib5" ref-type="bibr">Crossa <italic>et al.</italic> (2011)</xref> and <xref rid="bib15" ref-type="bibr">González-Camacho <italic>et al.</italic> (2012)</xref>, among others.</p></sec><sec id="s4"><title>Genotypic data:</title><p>Genotypes of all 278 lines were obtained using 53,401 SNPs. Those SNPs with >10% missing values or a minor allele frequency ≤0.05 were excluded from the data. Call rate per line was considered sufficient (>90%) for all lines. Most lines (250 of 278) had a call rate greater than 99%. After line-specific quality control (applying the same quality control to each line separately), the maize data still contained 46,347 SNPs, which were used in the analysis.</p></sec></sec><sec id="s5"><title>Statistical models</title><sec id="s6"><title>Threshold (cumulative probit) model:</title><p>Let <inline-formula><mml:math id="me11"><mml:mrow><mml:mi mathvariant="bold-italic">y</mml:mi><mml:mo>=</mml:mo><mml:mo>{</mml:mo><mml:msub><mml:mi>y</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>}</mml:mo></mml:mrow></mml:math></inline-formula> (<inline-formula><mml:math id="me12"><mml:mrow><mml:mi>i</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn><mml:mo>,</mml:mo><mml:mo>…</mml:mo><mml:mo>,</mml:mo><mml:mi>I</mml:mi><mml:mo>;</mml:mo><mml:mi>j</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn><mml:mo>,</mml:mo><mml:mn>2</mml:mn><mml:mo>,</mml:mo><mml:mo>…</mml:mo><mml:mo>,</mml:mo><mml:mi>J</mml:mi><mml:mo>;</mml:mo><mml:mi>k</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn><mml:mo>,</mml:mo><mml:mn>2</mml:mn><mml:mo>,</mml:mo><mml:mo>…</mml:mo><mml:msub><mml:mi>r</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>,</mml:mo></mml:mrow></mml:math></inline-formula> where <inline-formula><mml:math id="me13"><mml:mi>i</mml:mi></mml:math></inline-formula> represents the environment, <italic>j</italic> denotes the genotype and <italic>k</italic> is the number of replicates of each genotype in each environment. The response variable (disease resistance), <inline-formula><mml:math id="me14"><mml:mrow><mml:msub><mml:mi>y</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula>, represents an assignment into one of <inline-formula><mml:math id="me15"><mml:mi>C</mml:mi></mml:math></inline-formula> mutually exclusive and exhaustive categories (here <inline-formula><mml:math id="me16"><mml:mi>C</mml:mi></mml:math></inline-formula>=5, I = 3 and J = 278) that follow an order, since 1 indicates no infection, 2 low infection, 3 moderate infection, 4 high infection, and 5 complete infection. Therefore, in a GLMM framework, this model can be described by defining the distribution, the linear predictor and the link function.</p><sec id="s7"><title>Distribution:</title><p>There are two distributions, one for observations in the response variable: <inline-formula><mml:math id="me17"><mml:mrow><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>y</mml:mi><mml:mrow><mml:mn>1</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>y</mml:mi><mml:mrow><mml:mn>2</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>,</mml:mo><mml:mo>…</mml:mo><mml:mo>,</mml:mo><mml:msub><mml:mi>y</mml:mi><mml:mrow><mml:mi>C</mml:mi><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mrow><mml:mo>|</mml:mo><mml:mrow><mml:mi mathvariant="bold-italic">β</mml:mi><mml:mo>,</mml:mo><mml:mi mathvariant="bold-italic">b</mml:mi></mml:mrow></mml:mrow></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mo>∼</mml:mo><mml:mi mathvariant="normal">Multinomial</mml:mi><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>N</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>π</mml:mi><mml:mrow><mml:mn>1</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>π</mml:mi><mml:mrow><mml:mn>2</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>,</mml:mo><mml:mo>…</mml:mo><mml:mo>,</mml:mo><mml:msub><mml:mi>π</mml:mi><mml:mrow><mml:mi>C</mml:mi><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub></mml:mrow><mml:mo>)</mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula>, where <inline-formula><mml:math id="me18"><mml:mi mathvariant="bold-italic">β</mml:mi></mml:math></inline-formula> is the I×1 vector of fixed environmental effects and another distribution for the random effects, <inline-formula><mml:math id="me19"><mml:mrow><mml:mi mathvariant="bold-italic">b</mml:mi><mml:mo>∼</mml:mo><mml:mi mathvariant="normal">N</mml:mi><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:mi mathvariant="bold">0</mml:mi><mml:mo>,</mml:mo><mml:mi mathvariant="bold-italic">G</mml:mi></mml:mrow><mml:mo>)</mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula>, where <bold>G</bold> is the variance-covariance matrix of <inline-formula><mml:math id="me20"><mml:mrow><mml:mi mathvariant="bold-italic">b</mml:mi><mml:mo>=</mml:mo><mml:mrow><mml:mo>{</mml:mo><mml:mrow><mml:msub><mml:mi>b</mml:mi><mml:mi>j</mml:mi></mml:msub></mml:mrow><mml:mo>}</mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math id="me21"><mml:mrow><mml:msub><mml:mi>b</mml:mi><mml:mi>j</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the effect of line <inline-formula><mml:math id="me22"><mml:mi>j</mml:mi></mml:math></inline-formula>.</p></sec><sec id="s8"><title>Linear predictor:</title><p><inline-formula><mml:math id="me23"><mml:mrow><mml:msub><mml:mi>η</mml:mi><mml:mrow><mml:mi>c</mml:mi><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:msub><mml:mi>γ</mml:mi><mml:mi>c</mml:mi></mml:msub><mml:mo>−</mml:mo><mml:msubsup><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup><mml:mi mathvariant="bold-italic">β</mml:mi><mml:mo>−</mml:mo><mml:msubsup><mml:mi mathvariant="bold-italic">z</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup><mml:mi mathvariant="bold-italic">b</mml:mi></mml:mrow></mml:math></inline-formula>, where <inline-formula><mml:math id="me24"><mml:mrow><mml:msub><mml:mi>η</mml:mi><mml:mrow><mml:mi>c</mml:mi><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula> denotes the <inline-formula><mml:math id="me25"><mml:mi>c</mml:mi></mml:math></inline-formula><sup>th</sup> link (<inline-formula><mml:math id="me26"><mml:mrow><mml:mi>c</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn><mml:mo>,</mml:mo><mml:mn>2</mml:mn><mml:mo>,</mml:mo><mml:mo>…</mml:mo><mml:mo>,</mml:mo><mml:mi>C</mml:mi><mml:mo>−</mml:mo><mml:mn>1</mml:mn><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></inline-formula> for the fixed and random effects combination, <inline-formula><mml:math id="me27"><mml:mrow><mml:msub><mml:mi>γ</mml:mi><mml:mi>c</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the intercept (threshold) for the <inline-formula><mml:math id="me28"><mml:mi>c</mml:mi></mml:math></inline-formula><sup>th</sup> link, and <inline-formula><mml:math id="me29"><mml:mrow><mml:msubsup><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math id="me30"><mml:mrow><mml:msubsup><mml:mi mathvariant="bold-italic">z</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup></mml:mrow></mml:math></inline-formula> are known row incidence vectors corresponding to fixed and random effects in <inline-formula><mml:math id="me31"><mml:mi mathvariant="bold-italic">β</mml:mi></mml:math></inline-formula> and <inline-formula><mml:math id="me32"><mml:mi mathvariant="bold-italic">b</mml:mi></mml:math></inline-formula>, respectively. Since there are <inline-formula><mml:math id="me33"><mml:mi>C</mml:mi></mml:math></inline-formula> categories, a total of <inline-formula><mml:math id="me34"><mml:mrow><mml:mi>C</mml:mi><mml:mo>−</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:math></inline-formula> link functions are required to fully specify the model.</p></sec><sec id="s9"><title>Link function:</title><p>Cumulative probit <inline-formula><mml:math id="me35"><mml:mrow><mml:mo>{</mml:mo><mml:msub><mml:mi>η</mml:mi><mml:mrow><mml:mn>1</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:msup><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>−</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msup><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>π</mml:mi><mml:mrow><mml:mn>1</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mo>,</mml:mo><mml:msub><mml:mi>η</mml:mi><mml:mrow><mml:mn>2</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:msup><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>−</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msup><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>π</mml:mi><mml:mrow><mml:mn>1</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>π</mml:mi><mml:mrow><mml:mn>2</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mo>,</mml:mo><mml:mo>…</mml:mo><mml:mo>,</mml:mo><mml:msub><mml:mi>η</mml:mi><mml:mrow><mml:mo stretchy="false">(</mml:mo><mml:mi>C</mml:mi><mml:mo>−</mml:mo><mml:mn>1</mml:mn><mml:mo stretchy="false">)</mml:mo><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:msup><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>−</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msup><mml:mo stretchy="false">(</mml:mo><mml:msub><mml:mi>π</mml:mi><mml:mrow><mml:mn>1</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>π</mml:mi><mml:mrow><mml:mn>2</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:mo>…</mml:mo><mml:mo>+</mml:mo><mml:msub><mml:mi>π</mml:mi><mml:mrow><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:mi>C</mml:mi><mml:mo>−</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo stretchy="false">)</mml:mo><mml:mo>}</mml:mo></mml:mrow></mml:math></inline-formula>.<inline-formula><mml:math id="me36"><mml:mrow><mml:mi mathvariant="normal">Φ</mml:mi><mml:mrow><mml:mo>(</mml:mo><mml:mi mathvariant="normal">⋅</mml:mi><mml:mo>)</mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula> is the cumulative distribution function of a standard normal distribution (probit link) and <inline-formula><mml:math id="me37"><mml:mrow><mml:msup><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>−</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula> its corresponding inverse.</p><p>The inverse link for this model is given as follows: <inline-formula><mml:math id="me38"><mml:mrow><mml:msub><mml:mi>π</mml:mi><mml:mrow><mml:mn>1</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>η</mml:mi><mml:mrow><mml:mn>1</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mo>,</mml:mo><mml:msub><mml:mi>π</mml:mi><mml:mrow><mml:mn>1</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>π</mml:mi><mml:mrow><mml:mn>2</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>η</mml:mi><mml:mrow><mml:mn>2</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mo>,</mml:mo><mml:mo>…</mml:mo><mml:mo>,</mml:mo><mml:msub><mml:mi>π</mml:mi><mml:mrow><mml:mn>1</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>π</mml:mi><mml:mrow><mml:mn>2</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:mo>…</mml:mo><mml:mo>+</mml:mo><mml:msub><mml:mi>π</mml:mi><mml:mrow><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:mi>C</mml:mi><mml:mo>−</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>η</mml:mi><mml:mrow><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:mi>C</mml:mi><mml:mo>−</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub></mml:mrow><mml:mo>)</mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula>. Once we have estimates of <inline-formula><mml:math id="me39"><mml:mrow><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>η</mml:mi><mml:mrow><mml:mn>1</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mo>,</mml:mo><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>η</mml:mi><mml:mrow><mml:mn>2</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mo>,</mml:mo><mml:mo>…</mml:mo><mml:mo>,</mml:mo><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>η</mml:mi><mml:mrow><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:mi>C</mml:mi><mml:mo>−</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mo>,</mml:mo></mml:mrow></mml:math></inline-formula> we estimate <inline-formula><mml:math id="me40"><mml:mrow><mml:msub><mml:mover accent="true"><mml:mi>π</mml:mi><mml:mo>^</mml:mo></mml:mover><mml:mrow><mml:mn>2</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>η</mml:mi><mml:mrow><mml:mn>2</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mo>−</mml:mo><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>η</mml:mi><mml:mrow><mml:mn>1</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub></mml:mrow><mml:mo>)</mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula>, <inline-formula><mml:math id="me41"><mml:mrow><mml:msub><mml:mover accent="true"><mml:mi>π</mml:mi><mml:mo>^</mml:mo></mml:mover><mml:mrow><mml:mn>3</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>η</mml:mi><mml:mrow><mml:mn>3</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mo>−</mml:mo><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>η</mml:mi><mml:mrow><mml:mn>2</mml:mn><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub></mml:mrow><mml:mo>)</mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula>,…,<inline-formula><mml:math id="me42"><mml:mrow><mml:msub><mml:mover accent="true"><mml:mi>π</mml:mi><mml:mo>^</mml:mo></mml:mover><mml:mrow><mml:mi>C</mml:mi><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mn>1</mml:mn><mml:mo>−</mml:mo><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>η</mml:mi><mml:mrow><mml:mo stretchy="false">(</mml:mo><mml:mi>C</mml:mi><mml:mo>−</mml:mo><mml:mn>1</mml:mn><mml:mo stretchy="false">)</mml:mo><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub></mml:mrow><mml:mo>)</mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula>. This threshold model assumes that the process that gives rise to the observed categories is an underlying continuous variable with a normal distribution <inline-formula><mml:math id="me43"><mml:mrow><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:msubsup><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup><mml:mi mathvariant="bold-italic">β</mml:mi><mml:mo>+</mml:mo><mml:msubsup><mml:mi mathvariant="bold-italic">z</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup><mml:mi mathvariant="bold-italic">b</mml:mi><mml:mo>+</mml:mo><mml:msub><mml:mi>ε</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>,</mml:mo></mml:mrow></mml:math></inline-formula> where <inline-formula><mml:math id="me44"><mml:mrow><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula> are called “liabilities,” <inline-formula><mml:math id="me45"><mml:mrow><mml:msub><mml:mi>ε</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>∼</mml:mo><mml:mi>N</mml:mi><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:mn>0</mml:mn><mml:mo>,</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mo>)</mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula> (<italic>e.g.</italic>, <xref rid="bib13" ref-type="bibr">Gianola 1982</xref>, and <xref rid="bib29" ref-type="bibr">Sorensen <italic>et al.</italic> 1995</xref>). Furthermore, to generate ordinal categorical phenotypes with <inline-formula><mml:math id="me46"><mml:mi>C</mml:mi></mml:math></inline-formula> categories, the underlying phenotypic values are mapped to ordinal categorical phenotypes based on threshold parameters <inline-formula><mml:math id="me47"><mml:mrow><mml:msup><mml:mi mathvariant="bold-italic">γ</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:mo>=</mml:mo><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>γ</mml:mi><mml:mrow><mml:mi>m</mml:mi><mml:mi>i</mml:mi><mml:mi>n</mml:mi></mml:mrow></mml:msub><mml:mo><</mml:mo><mml:msub><mml:mi>γ</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:mo><</mml:mo><mml:mo>⋯</mml:mo><mml:mo><</mml:mo><mml:msub><mml:mi>γ</mml:mi><mml:mrow><mml:mi>C</mml:mi><mml:mo>−</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msub><mml:mo><</mml:mo><mml:msub><mml:mi>γ</mml:mi><mml:mrow><mml:mi>m</mml:mi><mml:mi>a</mml:mi><mml:mi>x</mml:mi></mml:mrow></mml:msub></mml:mrow><mml:mo>)</mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula> with <inline-formula><mml:math id="me48"><mml:mrow><mml:msub><mml:mi>γ</mml:mi><mml:mrow><mml:mi>m</mml:mi><mml:mi>i</mml:mi><mml:mi>n</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mo>−</mml:mo><mml:mn>∞</mml:mn></mml:mrow></mml:math></inline-formula>, and <inline-formula><mml:math id="me49"><mml:mrow><mml:msub><mml:mi>γ</mml:mi><mml:mrow><mml:mi>m</mml:mi><mml:mi>a</mml:mi><mml:mi>x</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mn>∞</mml:mn><mml:mo>,</mml:mo></mml:mrow></mml:math></inline-formula> which are <italic>cutpoints</italic> of the continuous scale such that the observed ordinal categorical response (<inline-formula><mml:math id="me50"><mml:mrow><mml:msub><mml:mi>y</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></inline-formula> is given by:<disp-formula id="eq"><mml:math id="me51"><mml:mrow><mml:msub><mml:mi>y</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mrow><mml:mo>{</mml:mo><mml:mrow><mml:mtable><mml:mtr><mml:mtd><mml:mrow><mml:mn>1</mml:mn><mml:mtext> </mml:mtext><mml:mi>i</mml:mi><mml:mi>f</mml:mi><mml:mo>−</mml:mo><mml:mn>∞</mml:mn><mml:mo><</mml:mo><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo><</mml:mo><mml:msub><mml:mi>γ</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:mo>,</mml:mo></mml:mrow></mml:mtd></mml:mtr><mml:mtr><mml:mtd><mml:mrow><mml:mn>2</mml:mn><mml:mtext> </mml:mtext><mml:mi>i</mml:mi><mml:mi>f</mml:mi><mml:mtext> </mml:mtext><mml:msub><mml:mi>γ</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:mo><</mml:mo><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo><</mml:mo><mml:msub><mml:mi>γ</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:mo>,</mml:mo></mml:mrow></mml:mtd></mml:mtr><mml:mtr><mml:mtd><mml:mo>⋮</mml:mo></mml:mtd></mml:mtr><mml:mtr><mml:mtd><mml:mrow><mml:mi>C</mml:mi><mml:mtext> </mml:mtext><mml:mi>i</mml:mi><mml:mi>f</mml:mi><mml:mtext> </mml:mtext><mml:msub><mml:mi>γ</mml:mi><mml:mrow><mml:mi>C</mml:mi><mml:mo>−</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msub><mml:mo><</mml:mo><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo><</mml:mo><mml:mn>∞</mml:mn></mml:mrow></mml:mtd></mml:mtr></mml:mtable></mml:mrow></mml:mrow></mml:mrow></mml:math></disp-formula>That is, <inline-formula><mml:math id="me52"><mml:mrow><mml:msub><mml:mi>y</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula> falls into category <inline-formula><mml:math id="me53"><mml:mi>C</mml:mi></mml:math></inline-formula> when the latent variable falls into the <italic>c</italic><sup>th</sup> interval of values. Given <inline-formula><mml:math id="me54"><mml:mi mathvariant="bold-italic">β</mml:mi></mml:math></inline-formula> and <inline-formula><mml:math id="me55"><mml:mi mathvariant="bold-italic">b</mml:mi></mml:math></inline-formula>, <inline-formula><mml:math id="me56"><mml:mrow><mml:mtext>the liabilities </mml:mtext><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula> are conditionally independent and distributed as</p><disp-formula id="eq___1"><mml:math id="me57"><mml:mrow><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mrow><mml:mo>|</mml:mo><mml:mrow><mml:mi mathvariant="bold-italic">β</mml:mi><mml:mo>,</mml:mo></mml:mrow></mml:mrow><mml:mi mathvariant="bold-italic">b</mml:mi><mml:mo>∼</mml:mo><mml:mi mathvariant="bold-italic">N</mml:mi><mml:mo stretchy="false">(</mml:mo><mml:msubsup><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup><mml:mi mathvariant="bold-italic">β</mml:mi><mml:mo>+</mml:mo><mml:msubsup><mml:mi mathvariant="bold-italic">z</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup><mml:mi mathvariant="bold-italic">b</mml:mi><mml:mo>,</mml:mo><mml:msubsup><mml:mi>σ</mml:mi><mml:mi>ε</mml:mi><mml:mn>2</mml:mn></mml:msubsup><mml:mo>=</mml:mo><mml:mn>1</mml:mn><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></disp-formula><p>where <inline-formula><mml:math id="me58"><mml:mrow><mml:msubsup><mml:mi>σ</mml:mi><mml:mi>ε</mml:mi><mml:mn>2</mml:mn></mml:msubsup></mml:mrow></mml:math></inline-formula> is fixed to 1 to achieve identifiability in the likelihood because the liabilities are unobservable. Also, one of the thresholds was fixed (<inline-formula><mml:math id="me59"><mml:mrow><mml:msub><mml:mi>γ</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:mo>=</mml:mo><mml:mn>0</mml:mn><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></inline-formula> to center the distribution. Therefore, <disp-formula id="eq___2"><mml:math id="me60"><mml:mrow><mml:mi>P</mml:mi><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>≤</mml:mo><mml:msub><mml:mi>γ</mml:mi><mml:mi>c</mml:mi></mml:msub><mml:mrow><mml:mo>|</mml:mo><mml:mrow><mml:mi mathvariant="bold-italic">β</mml:mi><mml:mo>,</mml:mo><mml:mi mathvariant="bold-italic">b</mml:mi></mml:mrow></mml:mrow></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mi mathvariant="normal">=</mml:mi><mml:mi mathvariant="bold">Φ</mml:mi><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>γ</mml:mi><mml:mi>c</mml:mi></mml:msub><mml:mo>−</mml:mo><mml:msubsup><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup><mml:mi mathvariant="bold-italic">β</mml:mi><mml:mo>−</mml:mo><mml:msubsup><mml:mi mathvariant="bold-italic">z</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup><mml:mi mathvariant="bold-italic">b</mml:mi></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mtext> </mml:mtext><mml:mi mathvariant="normal">for</mml:mi><mml:mtext> </mml:mtext><mml:mi>c</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn><mml:mo>,</mml:mo><mml:mn>2</mml:mn><mml:mo>,</mml:mo><mml:mo>…</mml:mo><mml:mo>,</mml:mo><mml:mi>C</mml:mi><mml:mo>−</mml:mo><mml:mn>1.</mml:mn></mml:mrow></mml:math></disp-formula>Note that <inline-formula><mml:math id="me61"><mml:mrow><mml:msub><mml:mi>γ</mml:mi><mml:mi>c</mml:mi></mml:msub><mml:mo>−</mml:mo><mml:msubsup><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup><mml:mi mathvariant="bold-italic">β</mml:mi><mml:mo>−</mml:mo><mml:msubsup><mml:mi mathvariant="bold-italic">z</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup><mml:mi mathvariant="bold-italic">b</mml:mi></mml:mrow></mml:math></inline-formula> is the predictor of the GLMM for the multinomial data given previously. Then, the conditional probability that <inline-formula><mml:math id="me62"><mml:mrow><mml:msub><mml:mi>y</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula> falls into category <inline-formula><mml:math id="me63"><mml:mi>c</mml:mi></mml:math></inline-formula> (<inline-formula><mml:math id="me64"><mml:mrow><mml:mi>c</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn><mml:mo>,</mml:mo><mml:mo>…</mml:mo><mml:mo>,</mml:mo><mml:mn>5</mml:mn><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></inline-formula> given <inline-formula><mml:math id="me65"><mml:mi mathvariant="bold-italic">β</mml:mi></mml:math></inline-formula>,<inline-formula><mml:math id="me66"><mml:mrow><mml:mi mathvariant="bold-italic">b</mml:mi><mml:mo>,</mml:mo></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math id="me67"><mml:mrow><mml:mi mathvariant="bold-italic">γ</mml:mi><mml:mo>=</mml:mo><mml:mo stretchy="false">(</mml:mo><mml:msub><mml:mi>γ</mml:mi><mml:mrow><mml:mi>m</mml:mi><mml:mi>i</mml:mi><mml:mi>n</mml:mi></mml:mrow></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>γ</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>γ</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>γ</mml:mi><mml:mrow><mml:mi>C</mml:mi><mml:mo>−</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>γ</mml:mi><mml:mrow><mml:mi>m</mml:mi><mml:mi>a</mml:mi><mml:mi>x</mml:mi></mml:mrow></mml:msub><mml:mo stretchy="false">)</mml:mo><mml:mo>,</mml:mo></mml:mrow></mml:math></inline-formula> is given by<disp-formula id="eq1"><mml:math id="me68"><mml:mrow><mml:mtable><mml:mtr><mml:mtd><mml:mrow><mml:mi>P</mml:mi><mml:mo stretchy="false">(</mml:mo><mml:msub><mml:mi>y</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mi>c</mml:mi><mml:mrow><mml:mo>|</mml:mo><mml:mrow><mml:mi mathvariant="bold-italic">β</mml:mi><mml:mo>,</mml:mo><mml:mi mathvariant="bold-italic">b</mml:mi><mml:mo>,</mml:mo><mml:mi mathvariant="bold-italic">γ</mml:mi><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:mrow></mml:mrow></mml:mtd><mml:mtd><mml:mrow><mml:mo>=</mml:mo><mml:mi>P</mml:mi><mml:mo stretchy="false">(</mml:mo><mml:msub><mml:mi>γ</mml:mi><mml:mrow><mml:mi>c</mml:mi><mml:mo>−</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msub><mml:mo><</mml:mo><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo><</mml:mo><mml:msub><mml:mi>γ</mml:mi><mml:mi>c</mml:mi></mml:msub><mml:mrow><mml:mo>|</mml:mo><mml:mrow><mml:mi mathvariant="bold-italic">β</mml:mi><mml:mo>,</mml:mo><mml:mi mathvariant="bold-italic">b</mml:mi><mml:mo>,</mml:mo><mml:mi mathvariant="bold-italic">γ</mml:mi><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:mrow></mml:mrow></mml:mtd></mml:mtr><mml:mtr><mml:mtd><mml:mrow></mml:mrow></mml:mtd><mml:mtd><mml:mrow><mml:mo>=</mml:mo><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>γ</mml:mi><mml:mi>c</mml:mi></mml:msub><mml:mo>−</mml:mo><mml:msubsup><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup><mml:mi mathvariant="bold-italic">β</mml:mi><mml:mo>−</mml:mo><mml:msubsup><mml:mi mathvariant="bold-italic">z</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup><mml:mi mathvariant="bold-italic">b</mml:mi></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mo>−</mml:mo><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>γ</mml:mi><mml:mrow><mml:mi>c</mml:mi><mml:mo>−</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msub><mml:mo>−</mml:mo><mml:msubsup><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup><mml:mi mathvariant="bold-italic">β</mml:mi><mml:mo>−</mml:mo><mml:msubsup><mml:mi mathvariant="bold-italic">z</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup><mml:mi mathvariant="bold-italic">b</mml:mi></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mo stretchy="false">(</mml:mo><mml:mn>1</mml:mn><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:mtd></mml:mtr></mml:mtable></mml:mrow></mml:math></disp-formula>The data are assumed conditionally independent, given <inline-formula><mml:math id="me69"><mml:mi mathvariant="bold-italic">β</mml:mi></mml:math></inline-formula>, <inline-formula><mml:math id="me70"><mml:mi mathvariant="bold-italic">b</mml:mi></mml:math></inline-formula>, and <inline-formula><mml:math id="me71"><mml:mi mathvariant="bold-italic">γ</mml:mi></mml:math></inline-formula>. Therefore, the sampling model is<disp-formula id="eq___3"><mml:math id="me72"><mml:mrow><mml:mtable columnalign="left"><mml:mtr columnalign="left"><mml:mtd columnalign="left"><mml:mrow><mml:mi>p</mml:mi><mml:mo stretchy="false">(</mml:mo><mml:mi mathvariant="bold-italic">y</mml:mi><mml:mrow><mml:mo>|</mml:mo><mml:mrow><mml:mi mathvariant="bold-italic">β</mml:mi><mml:mo>,</mml:mo><mml:mi mathvariant="bold-italic">b</mml:mi><mml:mo>,</mml:mo><mml:mi mathvariant="bold-italic">γ</mml:mi><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:mrow></mml:mrow></mml:mtd><mml:mtd columnalign="left"><mml:mrow><mml:mo>=</mml:mo><mml:munderover><mml:mstyle mathsize="140%" displaystyle="true"><mml:mo>∏</mml:mo></mml:mstyle><mml:mrow><mml:mi>i</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>I</mml:mi></mml:munderover><mml:munderover><mml:mstyle mathsize="140%" displaystyle="true"><mml:mo>∏</mml:mo></mml:mstyle><mml:mrow><mml:mi>j</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>J</mml:mi></mml:munderover><mml:munderover><mml:mstyle mathsize="140%" displaystyle="true"><mml:mo>∏</mml:mo></mml:mstyle><mml:mrow><mml:mi>k</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mrow><mml:msub><mml:mi>r</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:munderover><mml:munderover><mml:mstyle mathsize="140%" displaystyle="true"><mml:mo>∑</mml:mo></mml:mstyle><mml:mrow><mml:mi>c</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>C</mml:mi></mml:munderover><mml:msub><mml:mi>I</mml:mi><mml:mrow><mml:mrow><mml:mo>{</mml:mo><mml:mrow><mml:msub><mml:mi>y</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mi>c</mml:mi></mml:mrow><mml:mo>}</mml:mo></mml:mrow></mml:mrow></mml:msub><mml:mi>P</mml:mi><mml:mo stretchy="false">(</mml:mo><mml:msub><mml:mi>y</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>|</mml:mo><mml:mi mathvariant="bold-italic">β</mml:mi><mml:mo>,</mml:mo><mml:mi mathvariant="bold-italic">b</mml:mi><mml:mo>,</mml:mo><mml:mi mathvariant="bold-italic">γ</mml:mi><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:mtd></mml:mtr><mml:mtr columnalign="left"><mml:mtd columnalign="left"><mml:mrow></mml:mrow></mml:mtd><mml:mtd columnalign="left"><mml:mrow><mml:mo>=</mml:mo><mml:munderover><mml:mstyle mathsize="140%" displaystyle="true"><mml:mo>∏</mml:mo></mml:mstyle><mml:mrow><mml:mi>i</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>I</mml:mi></mml:munderover><mml:munderover><mml:mstyle mathsize="140%" displaystyle="true"><mml:mo>∏</mml:mo></mml:mstyle><mml:mrow><mml:mi>j</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>J</mml:mi></mml:munderover><mml:munderover><mml:mstyle mathsize="140%" displaystyle="true"><mml:mo>∏</mml:mo></mml:mstyle><mml:mrow><mml:mi>k</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mrow><mml:msub><mml:mi>r</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:munderover><mml:munderover><mml:mstyle mathsize="140%" displaystyle="true"><mml:mo>∑</mml:mo></mml:mstyle><mml:mrow><mml:mi>c</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>C</mml:mi></mml:munderover><mml:msub><mml:mi>I</mml:mi><mml:mrow><mml:mrow><mml:mo>{</mml:mo><mml:mrow><mml:msub><mml:mi>y</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mi>c</mml:mi></mml:mrow><mml:mo>}</mml:mo></mml:mrow></mml:mrow></mml:msub><mml:mo stretchy="false">[</mml:mo><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>γ</mml:mi><mml:mi>c</mml:mi></mml:msub><mml:mo>−</mml:mo><mml:msubsup><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup><mml:mi mathvariant="bold-italic">β</mml:mi><mml:mo>−</mml:mo><mml:msubsup><mml:mi mathvariant="bold-italic">z</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup><mml:mi mathvariant="bold-italic">b</mml:mi></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mo>−</mml:mo><mml:mtext>Φ</mml:mtext><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>γ</mml:mi><mml:mrow><mml:mi>c</mml:mi><mml:mo>−</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msub><mml:mo>−</mml:mo><mml:msubsup><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup><mml:mi mathvariant="bold-italic">β</mml:mi><mml:mo>−</mml:mo><mml:msubsup><mml:mi mathvariant="bold-italic">z</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow><mml:mi>T</mml:mi></mml:msubsup><mml:mi mathvariant="bold-italic">b</mml:mi></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mo stretchy="false">]</mml:mo></mml:mrow></mml:mtd></mml:mtr></mml:mtable></mml:mrow></mml:math></disp-formula>where <inline-formula><mml:math id="me73"><mml:mrow><mml:msub><mml:mi>I</mml:mi><mml:mrow><mml:mrow><mml:mo>{</mml:mo><mml:mrow><mml:msub><mml:mi>y</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mi>c</mml:mi></mml:mrow><mml:mo>}</mml:mo></mml:mrow></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula> is an indicator function taking a value of 1 if the response falls into category <inline-formula><mml:math id="me74"><mml:mrow><mml:mi>c</mml:mi><mml:mo>,</mml:mo></mml:mrow></mml:math></inline-formula> and 0 otherwise.</p></sec></sec><sec id="s10"><title>Models fitted:</title><p>Based on <xref rid="bib20" ref-type="bibr">Jarquín <italic>et al.</italic> (2014)</xref>, nine models are proposed for analyzing these data sets (<xref ref-type="table" rid="t1">Table 1</xref>). The sequence of models described below is similar to those presented by <xref rid="bib20" ref-type="bibr">Jarquín <italic>et al.</italic> (2014)</xref> for a continuous variable. However, for simplicity, only the underlying latent variables (<inline-formula><mml:math id="me75"><mml:mrow><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></inline-formula> are presented together with the distribution of the corresponding random effects that give rise to the observed categorical phenotypes, given that a probit link function is assumed by all models described below.</p><table-wrap id="t1" position="float"><label>Table 1</label><caption><title>Nine models used to fit the data set</title></caption><table frame="above" rules="groups"><col width="15.42%" span="1"></col><col width="15.53%" span="1"></col><col width="13.81%" span="1"></col><col width="13.81%" span="1"></col><col width="13.81%" span="1"></col><col width="13.81%" span="1"></col><col width="13.81%" span="1"></col><thead><tr><th rowspan="2" valign="top" align="left" scope="col" colspan="1">Model</th><th colspan="3" valign="top" align="center" scope="colgroup" rowspan="1">Main Effects</th><th colspan="3" valign="top" align="center" scope="colgroup" rowspan="1">Interaction</th></tr><tr><th valign="top" colspan="1" align="center" scope="colgroup" rowspan="1">E</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">L</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">G</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">G×G</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">G×E</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">G×G×E</th></tr></thead><tbody><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">1</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="left" rowspan="1" colspan="1"></td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">2</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="left" rowspan="1" colspan="1"></td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">3</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="left" rowspan="1" colspan="1"></td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">4</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="left" rowspan="1" colspan="1"></td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">5</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="left" rowspan="1" colspan="1"></td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">6</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="left" rowspan="1" colspan="1"></td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">7</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="center" rowspan="1" colspan="1">X</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">8</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="left" rowspan="1" colspan="1"></td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">9</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="center" rowspan="1" colspan="1">X</td><td valign="top" align="center" rowspan="1" colspan="1">X</td></tr></tbody></table><table-wrap-foot><fn><p>E, environment; L, line; G, marker covariates; G×G, additive × additive epistasis term; G×E, environment × marker interaction; G×G×E, additive × additive epistasis × environment interaction term.</p></fn></table-wrap-foot></table-wrap><sec id="s11"><title>Model 1:</title><p>The first model used to explain the liability value of the <italic>k</italic><sup>th</sup> individual in the <italic>j</italic><sup>th</sup> line at the <italic>i</italic><sup>th</sup> environment is<disp-formula id="eq2"><mml:math id="me76"><mml:mrow><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:msub><mml:mi>E</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>L</mml:mi><mml:mi>j</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>ε</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math><label>(2)</label></disp-formula>where <inline-formula><mml:math id="me77"><mml:mrow><mml:msub><mml:mi>E</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the fixed effect of the <italic>i</italic><sup>th</sup> environment and coefficient regressions were assigned a flat Gaussian prior with mean zero and variance equal to <inline-formula><mml:math id="me78"><mml:mrow><mml:mn>1</mml:mn><mml:mo>×</mml:mo><mml:msup><mml:mrow><mml:mn>10</mml:mn></mml:mrow><mml:mrow><mml:mn>10</mml:mn></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula>, <inline-formula><mml:math id="me79"><mml:mrow><mml:msub><mml:mi>L</mml:mi><mml:mi>j</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the random effect of the <italic>j</italic><sup>th</sup> line which is assumed to be identically and independently distributed (<italic>IID</italic>) as normal, <inline-formula><mml:math id="me80"><mml:mrow><mml:msub><mml:mi>L</mml:mi><mml:mi>j</mml:mi></mml:msub><mml:mover><mml:mo>∼</mml:mo><mml:mrow><mml:mi>I</mml:mi><mml:mi>I</mml:mi><mml:mi>D</mml:mi></mml:mrow></mml:mover><mml:mi>N</mml:mi><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:mn>0</mml:mn><mml:mo>,</mml:mo><mml:msubsup><mml:mi>σ</mml:mi><mml:mi>L</mml:mi><mml:mn>2</mml:mn></mml:msubsup></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mo>,</mml:mo></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math id="me81"><mml:mrow><mml:msub><mml:mi>ε</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula> is an error term distributed as <inline-formula><mml:math id="me82"><mml:mrow><mml:msub><mml:mi>ε</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mover><mml:mo>∼</mml:mo><mml:mrow><mml:mi>I</mml:mi><mml:mi>I</mml:mi><mml:mi>D</mml:mi></mml:mrow></mml:mover><mml:mi>N</mml:mi><mml:mo stretchy="false">(</mml:mo><mml:mn>0</mml:mn><mml:mo>,</mml:mo><mml:mn>1</mml:mn><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></inline-formula>. The unknown variance parameters (<inline-formula><mml:math id="me83"><mml:mrow><mml:msubsup><mml:mi>σ</mml:mi><mml:mi>L</mml:mi><mml:mn>2</mml:mn></mml:msubsup><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></inline-formula> were assigned scaled-inverted χ<sup>2</sup> distributions as prior. This density is indexed by two hyper-parameters: degrees of freedom (<italic>df</italic>) and the scale (<italic>S</italic>) parameter. In this case, we defined these values using the internal rules provided by BGLR. By default, that is, if the user does not specify these parameters, BGLR assigns mildly informative priors with <inline-formula><mml:math id="me84"><mml:mrow><mml:mi>d</mml:mi><mml:mi>f</mml:mi></mml:mrow></mml:math></inline-formula>=5 and a scale parameter that gives a prior mode, <inline-formula><mml:math id="me85"><mml:mrow><mml:mfrac><mml:mi>S</mml:mi><mml:mrow><mml:mi>d</mml:mi><mml:mi>f</mml:mi><mml:mo>+</mml:mo><mml:mn>2</mml:mn></mml:mrow></mml:mfrac></mml:mrow></mml:math></inline-formula>, that obeys a prior variance partition where 50% of the variance of the liability score corresponds to error terms, and the remaining 50% to the random effects included in the model. The internal rules implemented in BGLR are fully explained in <xref rid="bib25" ref-type="bibr">Pérez-Rodríguez and de los Campos (2014)</xref>.</p></sec><sec id="s12"><title>Model 2:</title><p>The second model is an extension of the GBLUP for ordered categorical phenotypes. This model was obtained from model 1 by replacing the line effect, <inline-formula><mml:math id="me86"><mml:mrow><mml:msub><mml:mi>L</mml:mi><mml:mi>j</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>, with a random effect that incorporates marker information using the genomic relationship matrix <bold><italic>G</italic></bold>. Model 2 expresses the liability value of the <italic>k</italic><sup>th</sup> individual in the <italic>j</italic><sup>th</sup> line in the <italic>i</italic><sup>th</sup> environment as<disp-formula id="eq3"><mml:math id="me87"><mml:mrow><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:msub><mml:mi>E</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>g</mml:mi><mml:mi>j</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>ε</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math><label>(3)</label></disp-formula>with <inline-formula><mml:math id="me88"><mml:mrow><mml:msub><mml:mi>g</mml:mi><mml:mi>j</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:munderover><mml:mstyle mathsize="140%" displaystyle="true"><mml:mo>∑</mml:mo></mml:mstyle><mml:mrow><mml:mi>m</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>p</mml:mi></mml:munderover><mml:msub><mml:mi>x</mml:mi><mml:mrow><mml:mi>j</mml:mi><mml:mi>m</mml:mi></mml:mrow></mml:msub><mml:msub><mml:mi>b</mml:mi><mml:mi>m</mml:mi></mml:msub><mml:mo>,</mml:mo></mml:mrow></mml:math></inline-formula> where <inline-formula><mml:math id="me89"><mml:mrow><mml:msub><mml:mi>g</mml:mi><mml:mi>j</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> represents an approximation to the true genetic value of the <italic>j</italic><sup>th</sup> line, <inline-formula><mml:math id="me90"><mml:mrow><mml:msub><mml:mi>x</mml:mi><mml:mrow><mml:mi>j</mml:mi><mml:mi>m</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula> is the genotype of the <italic>j</italic><sup>th</sup> line at the <italic>m</italic><sup>th</sup> marker (scored as 0 or 2 for genotypes that are homozygous at minor and major allele frequency, respectively, and 1 for heterozygous genotypes), and <inline-formula><mml:math id="me91"><mml:mrow><mml:msub><mml:mi>b</mml:mi><mml:mi>m</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the effect of the <italic>m</italic><sup>th</sup> marker with <inline-formula><mml:math id="me92"><mml:mrow><mml:msub><mml:mi>b</mml:mi><mml:mi>m</mml:mi></mml:msub><mml:mover><mml:mo>∼</mml:mo><mml:mrow><mml:mi>I</mml:mi><mml:mi>I</mml:mi><mml:mi>D</mml:mi></mml:mrow></mml:mover><mml:mi>N</mml:mi><mml:mo stretchy="false">(</mml:mo><mml:mn>0</mml:mn><mml:mo>,</mml:mo><mml:msubsup><mml:mi>σ</mml:mi><mml:mi>b</mml:mi><mml:mn>2</mml:mn></mml:msubsup><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></inline-formula>, <inline-formula><mml:math id="me93"><mml:mrow><mml:mo stretchy="false">(</mml:mo><mml:mi>m</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn><mml:mo>,</mml:mo><mml:mo>…</mml:mo><mml:mo>,</mml:mo><mml:mi>p</mml:mi><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></inline-formula>. Therefore, <inline-formula><mml:math id="me94"><mml:mrow><mml:mi mathvariant="bold-italic">g</mml:mi><mml:mo>=</mml:mo><mml:mi mathvariant="bold-italic">X</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">b</mml:mi><mml:mi mathvariant="bold-italic">m</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> contains the genomic values of all lines, and <inline-formula><mml:math id="me95"><mml:mrow><mml:mi mathvariant="bold-italic">g</mml:mi><mml:mo>=</mml:mo><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:msub><mml:mi>g</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:mo>,</mml:mo><mml:mo>…</mml:mo><mml:mo>,</mml:mo><mml:msub><mml:mi>g</mml:mi><mml:mrow><mml:mn>278</mml:mn></mml:mrow></mml:msub></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mi mathvariant="bold">′</mml:mi></mml:mrow></mml:math></inline-formula> is assumed to follow the normal distribution <inline-formula><mml:math id="me96"><mml:mrow><mml:mi mathvariant="bold-italic">g</mml:mi><mml:mo>∼</mml:mo><mml:mi>N</mml:mi><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:mn>0</mml:mn><mml:mo>,</mml:mo><mml:mi mathvariant="bold-italic">G</mml:mi><mml:msubsup><mml:mi>σ</mml:mi><mml:mi>g</mml:mi><mml:mn>2</mml:mn></mml:msubsup></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mo>,</mml:mo></mml:mrow></mml:math></inline-formula> where <inline-formula><mml:math id="me97"><mml:mi mathvariant="bold-italic">G</mml:mi></mml:math></inline-formula> is a marker-derived genomic relationship matrix that has an expected value equal to (under ideal conditions) twice the coefficient of parentage matrix of lines, and <inline-formula><mml:math id="me98"><mml:mrow><mml:msubsup><mml:mi>σ</mml:mi><mml:mi>g</mml:mi><mml:mn>2</mml:mn></mml:msubsup></mml:mrow></mml:math></inline-formula> is a genomic variance.</p><p>The entries of <inline-formula><mml:math id="me99"><mml:mi mathvariant="bold-italic">G</mml:mi></mml:math></inline-formula> were computed as <inline-formula><mml:math id="me100"><mml:mrow><mml:msub><mml:mi>G</mml:mi><mml:mrow><mml:mi>j</mml:mi><mml:msup><mml:mi>j</mml:mi><mml:mo>’</mml:mo></mml:msup></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:msup><mml:mi>p</mml:mi><mml:mrow><mml:mo>−</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msup><mml:munderover><mml:mstyle mathsize="140%" displaystyle="true"><mml:mo>∑</mml:mo></mml:mstyle><mml:mrow><mml:mi>m</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>p</mml:mi></mml:munderover><mml:msub><mml:mi>z</mml:mi><mml:mrow><mml:mi>j</mml:mi><mml:mi>m</mml:mi></mml:mrow></mml:msub><mml:mo stretchy="false">(</mml:mo><mml:msub><mml:mi>z</mml:mi><mml:mrow><mml:msup><mml:mi>j</mml:mi><mml:mo>’</mml:mo></mml:msup><mml:mi>m</mml:mi></mml:mrow></mml:msub><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></inline-formula>, where <inline-formula><mml:math id="me101"><mml:mrow><mml:msub><mml:mi>z</mml:mi><mml:mrow><mml:mi>j</mml:mi><mml:mi>m</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mo stretchy="false">(</mml:mo><mml:msub><mml:mi>x</mml:mi><mml:mrow><mml:mi>j</mml:mi><mml:mi>m</mml:mi></mml:mrow></mml:msub><mml:mo>−</mml:mo><mml:mn>2</mml:mn><mml:msub><mml:mi>p</mml:mi><mml:mi>m</mml:mi></mml:msub><mml:mo stretchy="false">)</mml:mo><mml:mo>/</mml:mo><mml:msqrt><mml:mrow><mml:mn>2</mml:mn><mml:msub><mml:mi>p</mml:mi><mml:mi>m</mml:mi></mml:msub><mml:mo stretchy="false">(</mml:mo><mml:mn>1</mml:mn><mml:mo>−</mml:mo><mml:msub><mml:mi>p</mml:mi><mml:mi>m</mml:mi></mml:msub><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:msqrt></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math id="me102"><mml:mrow><mml:msub><mml:mi>p</mml:mi><mml:mi>m</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the estimated frequency of the <italic>m</italic><sup>th</sup> marker. Subtracting <inline-formula><mml:math id="me103"><mml:mrow><mml:mn>2</mml:mn><mml:msub><mml:mi>p</mml:mi><mml:mi>m</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> from the genotype codes (centering) and dividing each marker covariate by <inline-formula><mml:math id="me104"><mml:mrow><mml:msqrt><mml:mrow><mml:mn>2</mml:mn><mml:msub><mml:mi>p</mml:mi><mml:mi>m</mml:mi></mml:msub><mml:mo stretchy="false">(</mml:mo><mml:mn>1</mml:mn><mml:mo>−</mml:mo><mml:msub><mml:mi>p</mml:mi><mml:mi>m</mml:mi></mml:msub><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:msqrt></mml:mrow></mml:math></inline-formula> (standardizing) is not strictly needed; however, standardization allows interpreting <inline-formula><mml:math id="me105"><mml:mrow><mml:msubsup><mml:mi>σ</mml:mi><mml:mi>g</mml:mi><mml:mn>2</mml:mn></mml:msubsup></mml:mrow></mml:math></inline-formula> as a genomic variance. <inline-formula><mml:math id="me106"><mml:mi mathvariant="bold-italic">G</mml:mi></mml:math></inline-formula> was constructed using the genotypes of all 278 lines.</p></sec><sec id="s13"><title>Model 3:</title><p>The third model is an extension of model 2 that accounts for “marked” epistatic additive × additive relationships, <inline-formula><mml:math id="me107"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">g</mml:mi><mml:mi mathvariant="bold-italic">A</mml:mi></mml:msub><mml:mo>∼</mml:mo><mml:mi>N</mml:mi><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:mn>0</mml:mn><mml:mo>,</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">G</mml:mi><mml:mi mathvariant="bold-italic">A</mml:mi></mml:msub><mml:msubsup><mml:mi>σ</mml:mi><mml:mrow><mml:mi>A</mml:mi><mml:mi>g</mml:mi></mml:mrow><mml:mn>2</mml:mn></mml:msubsup></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mo>,</mml:mo></mml:mrow></mml:math></inline-formula> where <inline-formula><mml:math id="me108"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">g</mml:mi><mml:mi mathvariant="bold-italic">A</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> represents a regression on genomic epistatic additive × additive relationships with <inline-formula><mml:math id="me109"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">G</mml:mi><mml:mi mathvariant="bold-italic">A</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:mi mathvariant="bold-italic">G</mml:mi><mml:mo>#</mml:mo><mml:mi mathvariant="bold-italic">G</mml:mi></mml:mrow></mml:math></inline-formula> (# is the element-wise multiplication operator, that is, a Hadamard product). The prior used for <inline-formula><mml:math id="me110"><mml:mrow><mml:msubsup><mml:mi>σ</mml:mi><mml:mrow><mml:mi>A</mml:mi><mml:mi>g</mml:mi></mml:mrow><mml:mn>2</mml:mn></mml:msubsup></mml:mrow></mml:math></inline-formula> was a scaled-inverted χ<sup>2</sup> distribution. In this model, the liability score of the <italic>k</italic><sup>th</sup> individual in the <italic>j</italic><sup>th</sup> line in the <italic>i</italic><sup>th</sup> environment is equal to</p><disp-formula id="eq4"><mml:math id="me111"><mml:mrow><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:msub><mml:mi>E</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>g</mml:mi><mml:mi>j</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>g</mml:mi><mml:mrow><mml:mi>A</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>ε</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math><label>(4)</label></disp-formula></sec><sec id="s14"><title>Model 4:</title><p>The third model combines model 1 and model 2 as <disp-formula id="eq5"><mml:math id="me112"><mml:mrow><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:msub><mml:mi>E</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>L</mml:mi><mml:mi>j</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>g</mml:mi><mml:mi>j</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>ε</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math><label>(5)</label></disp-formula>Model 4 partitions the line effects into two components, one that is explained by a regression on markers (<inline-formula><mml:math id="me113"><mml:mrow><mml:msub><mml:mi>g</mml:mi><mml:mi>j</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>) and another representing variation among lines that is not explained by regression on markers.</p></sec><sec id="s15"><title>Model 5:</title><p>This model is an extension of model 4 that accounts for epistatic additive × additive relationships, with liability equal to<disp-formula id="eq6"><mml:math id="me114"><mml:mrow><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:msub><mml:mi>E</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>L</mml:mi><mml:mi>j</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>g</mml:mi><mml:mi>j</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>g</mml:mi><mml:mrow><mml:mi>A</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>ε</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math><label>(6)</label></disp-formula>Note that none of the aforementioned models accounts for G×E (<inline-formula><mml:math id="me115"><mml:mrow><mml:mi>g</mml:mi><mml:msub><mml:mi>E</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></inline-formula> and G×G×E (<inline-formula><mml:math id="me116"><mml:mrow><mml:msub><mml:mi>g</mml:mi><mml:mi>A</mml:mi></mml:msub><mml:msub><mml:mi>E</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></inline-formula> interactions. Next, we consider models that incorporate G×E and G×G×E.</p></sec><sec id="s16"><title>Model 6:</title><p>Model 6 considers model 2 but, in addition to the main effects of environments (E) and markers (G), it takes into account the interaction between markers and environments (G×E). Because the data include multiple phenotypic records per line, the genetic covariance structure of genetic effects in the full data set is equal to <inline-formula><mml:math id="me117"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">Z</mml:mi><mml:mi>g</mml:mi></mml:msub><mml:mi mathvariant="bold-italic">G</mml:mi><mml:msubsup><mml:mi mathvariant="bold-italic">Z</mml:mi><mml:mi>g</mml:mi><mml:mi>T</mml:mi></mml:msubsup></mml:mrow></mml:math></inline-formula>, where <inline-formula><mml:math id="me118"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">Z</mml:mi><mml:mi>g</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the incidence matrix for the vector of additive genetic effects of markers (<xref rid="bib20" ref-type="bibr">Jarquín <italic>et al.</italic> 2014</xref>). Therefore, the covariance structure for the vector of interaction terms in the full data set <inline-formula><mml:math id="me119"><mml:mrow><mml:mi mathvariant="bold-italic">gE</mml:mi><mml:mo>=</mml:mo><mml:mrow><mml:mo>{</mml:mo><mml:mrow><mml:mi>g</mml:mi><mml:msub><mml:mi>E</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub></mml:mrow><mml:mo>}</mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula> is the Hadamard product of <inline-formula><mml:math id="me120"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">Z</mml:mi><mml:mi>g</mml:mi></mml:msub><mml:mi mathvariant="bold-italic">G</mml:mi><mml:msubsup><mml:mi mathvariant="bold-italic">Z</mml:mi><mml:mi>g</mml:mi><mml:mi>T</mml:mi></mml:msubsup></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math id="me121"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">Z</mml:mi><mml:mi>E</mml:mi></mml:msub><mml:msubsup><mml:mi mathvariant="bold-italic">Z</mml:mi><mml:mi>E</mml:mi><mml:mi>T</mml:mi></mml:msubsup><mml:mo>,</mml:mo></mml:mrow></mml:math></inline-formula> where <inline-formula><mml:math id="me122"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">Z</mml:mi><mml:mi>E</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> represents the incidence matrix of the effects of environments (<italic>i.e.</italic>, the matrix that connects phenotypes with environments) (<xref rid="bib20" ref-type="bibr">Jarquín <italic>et al.</italic> 2014</xref>). Therefore, the liability value of the <italic>k</italic><sup>th</sup> individual in the <italic>j</italic><sup>th</sup> line in the <italic>i</italic><sup>th</sup> environment is explained by <disp-formula id="eq7"><mml:math id="me123"><mml:mrow><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:msub><mml:mi>E</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>g</mml:mi><mml:mi>j</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mi>g</mml:mi><mml:msub><mml:mi>E</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>ε</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math><label>(7)</label></disp-formula>where <inline-formula><mml:math id="me124"><mml:mrow><mml:mi mathvariant="bold-italic">g</mml:mi><mml:mo>∼</mml:mo><mml:mi>N</mml:mi><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:mn>0</mml:mn><mml:mo>,</mml:mo><mml:mi mathvariant="bold-italic">G</mml:mi><mml:msubsup><mml:mi>σ</mml:mi><mml:mi>g</mml:mi><mml:mn>2</mml:mn></mml:msubsup></mml:mrow><mml:mo>)</mml:mo></mml:mrow><mml:mo>,</mml:mo><mml:mi mathvariant="bold-italic">gE</mml:mi><mml:mo>∼</mml:mo><mml:mi>N</mml:mi><mml:mo stretchy="false">(</mml:mo><mml:mn>0</mml:mn><mml:mo>,</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">Z</mml:mi><mml:mi>g</mml:mi></mml:msub><mml:mi mathvariant="bold-italic">G</mml:mi><mml:msubsup><mml:mi mathvariant="bold-italic">Z</mml:mi><mml:mi>g</mml:mi><mml:mi>T</mml:mi></mml:msubsup><mml:mo>#</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">Z</mml:mi><mml:mi>E</mml:mi></mml:msub><mml:msubsup><mml:mi mathvariant="bold-italic">Z</mml:mi><mml:mi>E</mml:mi><mml:mi>T</mml:mi></mml:msubsup><mml:msubsup><mml:mi>σ</mml:mi><mml:mrow><mml:mi>g</mml:mi><mml:mi>E</mml:mi></mml:mrow><mml:mn>2</mml:mn></mml:msubsup><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math id="me125"><mml:mrow><mml:msub><mml:mi>ε</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mover><mml:mo>∼</mml:mo><mml:mrow><mml:mi>I</mml:mi><mml:mi>I</mml:mi><mml:mi>D</mml:mi></mml:mrow></mml:mover><mml:mi>N</mml:mi><mml:mrow><mml:mo>(</mml:mo><mml:mrow><mml:mn>0</mml:mn><mml:mo>,</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mo>)</mml:mo></mml:mrow></mml:mrow></mml:math></inline-formula>.</p></sec><sec id="s17"><title>Model 7:</title><p>Model 7 extends model 6 to account for marked epistatic additive × additive relationships and the interaction between the epistatic additive × additive term and the environments. The liability value of the <italic>k</italic><sup>th</sup> individual in the <italic>j</italic><sup>th</sup> line in the <italic>i</italic><sup>th</sup> environment is now <disp-formula id="eq8"><mml:math id="me126"><mml:mrow><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:msub><mml:mi>E</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>g</mml:mi><mml:mi>j</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>g</mml:mi><mml:mrow><mml:mi>A</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:mi>g</mml:mi><mml:msub><mml:mi>E</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>g</mml:mi><mml:mi>A</mml:mi></mml:msub><mml:msub><mml:mi>E</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>ε</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math><label>(8)</label></disp-formula>where <inline-formula><mml:math id="me127"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">g</mml:mi><mml:mi mathvariant="bold-italic">A</mml:mi></mml:msub><mml:mi mathvariant="bold-italic">E</mml:mi><mml:mo>∼</mml:mo><mml:mi>N</mml:mi><mml:mo stretchy="false">(</mml:mo><mml:mn>0</mml:mn><mml:mo>,</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">Z</mml:mi><mml:mi>g</mml:mi></mml:msub><mml:msub><mml:mi mathvariant="bold-italic">G</mml:mi><mml:mi mathvariant="bold-italic">A</mml:mi></mml:msub><mml:msubsup><mml:mi mathvariant="bold-italic">Z</mml:mi><mml:mi>g</mml:mi><mml:mi>T</mml:mi></mml:msubsup><mml:mo>#</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">Z</mml:mi><mml:mi>E</mml:mi></mml:msub><mml:msubsup><mml:mi mathvariant="bold-italic">Z</mml:mi><mml:mi>E</mml:mi><mml:mi>T</mml:mi></mml:msubsup><mml:msubsup><mml:mi>σ</mml:mi><mml:mrow><mml:msub><mml:mi>g</mml:mi><mml:mi>A</mml:mi></mml:msub><mml:mi>E</mml:mi></mml:mrow><mml:mn>2</mml:mn></mml:msubsup><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></inline-formula> with <inline-formula><mml:math id="me128"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">G</mml:mi><mml:mi mathvariant="bold-italic">A</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:mi mathvariant="bold-italic">G</mml:mi><mml:mo>#</mml:mo><mml:mi mathvariant="bold-italic">G</mml:mi></mml:mrow></mml:math></inline-formula>.</p></sec><sec id="s18"><title>Model 8:</title><p>Model 8 extends model 6 by adding the random effect of the lines (<inline-formula><mml:math id="me129"><mml:mrow><mml:msub><mml:mi>L</mml:mi><mml:mi>j</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>). Thus, the liability value of the <italic>k</italic><sup>th</sup> individual in the <italic>j</italic><sup>th</sup> line in the <italic>i</italic><sup>th</sup> environment is explained by</p><disp-formula id="eq9"><mml:math id="me130"><mml:mrow><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:msub><mml:mi>E</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>L</mml:mi><mml:mi>j</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>g</mml:mi><mml:mi>j</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mi>g</mml:mi><mml:msub><mml:mi>E</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>ε</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math><label>(9)</label></disp-formula></sec><sec id="s19"><title>Model 9:</title><p>Finally, model 9 extends model 8 by adding marked epistatic additive × additive relationships and the interaction between the epistatic additive × additive term and the environments. The liability value of the <italic>k</italic><sup>th</sup> individual in the <italic>j</italic><sup>th</sup> line in the <italic>i</italic><sup>th</sup> environment is explained by </p><disp-formula id="eq10"><mml:math id="me131"><mml:mrow><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:msub><mml:mi>E</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>L</mml:mi><mml:mi>j</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>g</mml:mi><mml:mi>j</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>g</mml:mi><mml:mrow><mml:mi>A</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:mi>g</mml:mi><mml:msub><mml:mi>E</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>g</mml:mi><mml:mi>A</mml:mi></mml:msub><mml:msub><mml:mi>E</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi>ε</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math><label>(10)</label></disp-formula></sec></sec><sec id="s20"><title>Implementation with BGLR:</title><p>The nine models were fitted using the R-package BGLR (<xref rid="bib8" ref-type="bibr">de los Campos and Pérez-Rodríguez 2013</xref>) in the R-software (<xref rid="bib28" ref-type="bibr">R Core Team 2014</xref>). Parametric and semi-parametric models with both molecular marker and pedigree data can be fitted in this R package. It also allows including various random effects with user-defined covariance matrices. In <xref ref-type="app" rid="app1">Appendix 1</xref>, we provide the R code used for fitting the most parameterized model (model 9). The implementation of the other models is similar to that of model 9 but with fewer random effects. All these models are implemented via a Bayesian approach using the Gibbs sampler algorithm, and sampling from the fully conditional distributions, as shown in <xref rid="bib29" ref-type="bibr">Sorensen <italic>et al.</italic> (1995)</xref>. For more details about Gibbs sampler implementation used in the R-package BGLR (<xref rid="bib8" ref-type="bibr">de los Campos and Pérez-Rodríguez 2013</xref>) for binary and ordered categorical phenotypes, refer to <xref rid="bib29" ref-type="bibr">Sorensen <italic>et al.</italic> (1995)</xref>. The prior distributions of the variance components of all the models are described in the appendix of the BGLR package (<xref rid="bib8" ref-type="bibr">de los Campos and Pérez-Rodríguez 2013</xref>).</p></sec><sec id="s21"><title>Assessing predictive ability for ordered categorical phenotypic data:</title><p>The performance of the models was evaluated employing a cross-validation approach to estimate their prediction accuracies. To that end, we split each of the data sets into two parts (training and validation sets), where the training set was used to fit the model and the validation set was used to evaluate the training model’s prediction ability. A total of 20 random partitions were performed, and the validation results were averaged over the 20 partitions with observations assigned to training and testing completely at random. This represents a prediction problem similar to that labeled as CV2 in <xref rid="bib3" ref-type="bibr">Burgueño <italic>et al.</italic> (2012)</xref>, where the performance of some lines was observed in some environments (training set) but not in others (testing set).</p><p>Measuring prediction accuracy for categorical traits is more challenging than for quantitative traits, where the Euclidean metric appears as a natural choice. A variety of scoring rules have been proposed to assess accuracy for categorical traits. A scoring rule provides a summary measure for evaluating probabilistic prediction by assigning a numerical score based on the predictive distribution and on the event or value that materializes (<xref rid="bib11" ref-type="bibr">Garthwaite <italic>et al.</italic> 2005</xref>; <xref rid="bib22" ref-type="bibr">Kneib <italic>et al.</italic> 2007</xref>). This goal is achieved in a reasonable way by taking into account not only point prediction but the whole predictive distribution.</p><p>In the case of a multinomial model, this predictive distribution is simply obtained by computing the probabilities for all <inline-formula><mml:math id="me132"><mml:mrow><mml:mi>C</mml:mi></mml:mrow></mml:math></inline-formula> categories of the response according to the estimated model, <italic>i.e.</italic>, we obtain the predictive distribution <inline-formula><mml:math id="me133"><mml:mrow><mml:msub><mml:mrow><mml:mover accent="true"><mml:mi mathvariant="bold-italic">π</mml:mi><mml:mo stretchy="true">^</mml:mo></mml:mover></mml:mrow><mml:mi>i</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:mo stretchy="false">(</mml:mo><mml:msub><mml:mrow><mml:mover accent="true"><mml:mi>π</mml:mi><mml:mo stretchy="true">^</mml:mo></mml:mover></mml:mrow><mml:mrow><mml:mi>i</mml:mi><mml:mn>1</mml:mn></mml:mrow></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mrow><mml:mover accent="true"><mml:mi>π</mml:mi><mml:mo stretchy="true">^</mml:mo></mml:mover></mml:mrow><mml:mrow><mml:mi>i</mml:mi><mml:mn>2</mml:mn></mml:mrow></mml:msub><mml:mo>,</mml:mo><mml:mo>…</mml:mo><mml:mo>,</mml:mo><mml:msub><mml:mrow><mml:mover accent="true"><mml:mi>π</mml:mi><mml:mo stretchy="true">^</mml:mo></mml:mover></mml:mrow><mml:mrow><mml:mi>i</mml:mi><mml:mi>C</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula>), derived from the estimated model for observation <italic>i</italic>, and <inline-formula><mml:math id="me134"><mml:mrow><mml:msub><mml:mi>c</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the realized value for this observation in the data set. A scoring rule is any real-value function <inline-formula><mml:math id="me135"><mml:mrow><mml:mi>S</mml:mi><mml:mo stretchy="false">(</mml:mo><mml:msub><mml:mover accent="true"><mml:mi>π</mml:mi><mml:mo>^</mml:mo></mml:mover><mml:mrow><mml:mi>i</mml:mi><mml:mi>c</mml:mi></mml:mrow></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>c</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></inline-formula>, <inline-formula><mml:math id="me136"><mml:mrow><mml:msub><mml:mi>c</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:mn>1</mml:mn><mml:mo>,</mml:mo><mml:mn>2</mml:mn><mml:mo>,</mml:mo><mml:mo>…</mml:mo><mml:mo>,</mml:mo><mml:mi>C</mml:mi></mml:mrow></mml:math></inline-formula>, that assigns a value to the event that category <inline-formula><mml:math id="me137"><mml:mi>c</mml:mi></mml:math></inline-formula> is observed when <inline-formula><mml:math id="me138"><mml:mrow><mml:msub><mml:mover accent="true"><mml:mi>π</mml:mi><mml:mo>^</mml:mo></mml:mover><mml:mrow><mml:mi>i</mml:mi><mml:mi>c</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula> is the predictive probability for individual <inline-formula><mml:math id="me139"><mml:mi>i</mml:mi></mml:math></inline-formula> in category <inline-formula><mml:math id="me140"><mml:mi>c</mml:mi></mml:math></inline-formula>. A suitable score is the sum<disp-formula id="eq___4"><mml:math id="me141"><mml:mrow><mml:mi>S</mml:mi><mml:mo>=</mml:mo><mml:munderover><mml:mstyle mathsize="140%" displaystyle="true"><mml:mo>∑</mml:mo></mml:mstyle><mml:mrow><mml:mi>i</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>n</mml:mi></mml:munderover><mml:mi>S</mml:mi><mml:mo stretchy="false">(</mml:mo><mml:msub><mml:mover accent="true"><mml:mi>π</mml:mi><mml:mo>^</mml:mo></mml:mover><mml:mrow><mml:mi>i</mml:mi><mml:mi>c</mml:mi></mml:mrow></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi>c</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></disp-formula>where <inline-formula><mml:math id="me142"><mml:mi>S</mml:mi></mml:math></inline-formula> is the sum over all observations in the test data set. The hit rate (<italic>i.e.</italic>, the percentage of true positive predictions) and the log-likelihood are two popular scoring rules. However, both have the drawback that they involve only one of the probabilities of the predictive distribution (<inline-formula><mml:math id="me143"><mml:mrow><mml:msub><mml:mrow><mml:mover accent="true"><mml:mi mathvariant="bold-italic">π</mml:mi><mml:mo stretchy="true">^</mml:mo></mml:mover></mml:mrow><mml:mi>i</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>). In addition, it has been well documented that the log-likelihood score is sensitive to extreme observations. In our applications, we utilize the Brier score (<xref rid="bib2" ref-type="bibr">Brier 1950</xref>), which is equal to<disp-formula id="eq11"><mml:math id="me144"><mml:mrow><mml:mi>B</mml:mi><mml:mi>S</mml:mi><mml:mo>=</mml:mo><mml:msup><mml:mi>n</mml:mi><mml:mrow><mml:mo>−</mml:mo><mml:mn>1</mml:mn></mml:mrow></mml:msup><mml:munderover><mml:mstyle mathsize="140%" displaystyle="true"><mml:mo>∑</mml:mo></mml:mstyle><mml:mrow><mml:mi>i</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>n</mml:mi></mml:munderover><mml:munderover><mml:mstyle mathsize="140%" displaystyle="true"><mml:mo>∑</mml:mo></mml:mstyle><mml:mrow><mml:mi>c</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>C</mml:mi></mml:munderover><mml:msup><mml:mrow><mml:mo stretchy="false">(</mml:mo><mml:msub><mml:mover accent="true"><mml:mi>π</mml:mi><mml:mo>^</mml:mo></mml:mover><mml:mrow><mml:mi>i</mml:mi><mml:mi>c</mml:mi></mml:mrow></mml:msub><mml:mo>−</mml:mo><mml:msub><mml:mi>d</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>c</mml:mi></mml:mrow></mml:msub><mml:mo stretchy="false">)</mml:mo></mml:mrow><mml:mn>2</mml:mn></mml:msup></mml:mrow></mml:math><label>(11)</label></disp-formula>where <inline-formula><mml:math id="me145"><mml:mrow><mml:msub><mml:mi>d</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>c</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula> takes the value of 1 if the ordinal categorical response observed for individual <inline-formula><mml:math id="me146"><mml:mi>i</mml:mi></mml:math></inline-formula> falls into category <inline-formula><mml:math id="me147"><mml:mi>c</mml:mi></mml:math></inline-formula>, and <inline-formula><mml:math id="me148"><mml:mrow><mml:msub><mml:mi>d</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>c</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mn>0</mml:mn><mml:mtext> otherwise</mml:mtext></mml:mrow></mml:math></inline-formula>. This scoring rule uses all the information contained in the predictive distribution, not just a small part such as the hit rate or the log-likelihood score. Therefore, it is a reasonable choice for comparing categorical regression models, even though there are other scoring rules that also have good properties. The range of <inline-formula><mml:math id="me149"><mml:mrow><mml:mi>B</mml:mi><mml:mi>S</mml:mi></mml:mrow></mml:math></inline-formula> in <xref ref-type="disp-formula" rid="eq11">equation (11)</xref> is between 0 and 2. For this reason, we took <inline-formula><mml:math id="me150"><mml:mrow><mml:mi>B</mml:mi><mml:mi>S</mml:mi><mml:mo>/</mml:mo><mml:mn>2</mml:mn><mml:mo>,</mml:mo></mml:mrow></mml:math></inline-formula> to get the Brier score bound between 0 and 1, and lower scores imply better predictions.</p></sec><sec id="s22"><title>Data and software:</title><p>The phenotypic data for GLS in three environments (Mexico, Zimbabwe, and Colombia) for the 278 maize lines, the 46,347 SNPs data and the R scripts developed to fit the predictive models used in this study are given in the GLScode.rar file deposited at <ext-link ext-link-type="uri" xlink:href="http://repository.cimmyt.org/xmlui/handle/10883/4128">http://repository.cimmyt.org/xmlui/handle/10883/4128</ext-link>. The BGLR package (<xref rid="bib8" ref-type="bibr">de los Campos and Pérez-Rodríguez 2013</xref>) can be downloaded from CRAN.</p></sec></sec></sec><sec sec-type="results" id="s23"><title>Results</title><p><xref ref-type="fig" rid="fig1">Figure 1</xref> shows the relative frequencies of each category for the whole data set and for each country. The whole data set contains 2798 observations; category 3 has the most observations (923), and category 1 is the one with the fewest (234). This pattern was also observed in Zimbabwe (1485 total observations; 37 in category 1 and 581 in category 3) but somewhat more pronounced. Colombia had 832 observations, with 215 in category 3, 208 in category 2, and only 63 in category 5. Mexico had 481 observations, most of them in category 2 (212), and category 1 had the fewest data (26 observations).</p><fig id="fig1" fig-type="figure" position="float"><label>Figure 1</label><caption><p>Relative frequency of each category in the whole data set.</p></caption><graphic xlink:href="291f1"></graphic></fig><p>When applying eigenvalue decomposition to the genomic relationship matrix, 82 eigenvectors (components) of a total of 278 were needed to explain 80% of the total variance of genotypes. The first eigenvector captured 13.68% of the total variation of marker genotypes, whereas the second eigenvector explained 6.28% of the total variation. For these reasons, we can say that there is some, but not strong, evidence of population (and family) stratification, indicating a relatively diverse set of lines. This was expected because these lines came from different breeding programs.</p><p><xref ref-type="table" rid="t2">Table 2</xref> gives the estimates of the fixed (environment) effects and threshold parameters for each of the nine models; the first five models (1−5) had similar parameter estimates (fixed and threshold) that were different from those of models 6−9. Models 6−9 produced similar fixed and threshold parameter estimates. In plant and animal breeding, the focus is on estimating response probabilities associated with specific linear parameter combinations (<xref rid="bib14" ref-type="bibr">Gianola and Foulley 1983</xref>). <xref ref-type="fig" rid="fig2">Figure 2</xref> gives the probabilities for each ordinal categorical phenotype for model 9, for the whole data set, and for each country. In <xref ref-type="fig" rid="fig2">Figure 2</xref>, the average probabilities for category 5 (complete infection) were slightly greater than 0.20 (20%) in the whole data set and in each country. Colombia shows the greatest probability of complete infection (category 5), but it was only slightly greater than the probabilities in Mexico and Zimbabwe.</p><table-wrap id="t2" position="float"><label>Table 2</label><caption><title>Mean and SD of posterior distributions of fixed (environment) effects <inline-formula><mml:math id="me151"><mml:mrow><mml:mo mathvariant="bold">(</mml:mo><mml:msub><mml:mi>β</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:mtext> for Colombia</mml:mtext><mml:mo>,</mml:mo><mml:msub><mml:mi>β</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:mtext> for Zimbawe</mml:mtext><mml:mo>,</mml:mo><mml:mtext>and </mml:mtext><mml:msub><mml:mi>β</mml:mi><mml:mrow><mml:mn>3</mml:mn></mml:mrow></mml:msub><mml:mtext> for Mexico</mml:mtext><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></inline-formula> and threshold parameters of the nine proposed models</title></caption><table frame="above" rules="groups"><col width="9.02%" span="1"></col><col width="12.66%" span="1"></col><col width="12.43%" span="1"></col><col width="12.43%" span="1"></col><col width="13.13%" span="1"></col><col width="13.13%" span="1"></col><col width="13.36%" span="1"></col><col width="13.84%" span="1"></col><thead><tr><th rowspan="2" valign="top" align="left" scope="col" colspan="1">
Model</th><th colspan="3" valign="top" align="center" scope="colgroup" rowspan="1">Mean Fixed Parameters</th><th colspan="4" valign="top" align="center" scope="colgroup" rowspan="1"> Mean Threshold Parameters</th></tr><tr><th valign="top" colspan="1" align="center" scope="colgroup" rowspan="1"><inline-formula><mml:math id="me152"><mml:mrow><mml:msub><mml:mi>β</mml:mi><mml:mn>1</mml:mn></mml:msub></mml:mrow></mml:math></inline-formula></th><th valign="top" align="center" scope="col" rowspan="1" colspan="1"><inline-formula><mml:math id="me153"><mml:mrow><mml:msub><mml:mi>β</mml:mi><mml:mn>2</mml:mn></mml:msub></mml:mrow></mml:math></inline-formula></th><th valign="top" align="center" scope="col" rowspan="1" colspan="1"><inline-formula><mml:math id="me154"><mml:mrow><mml:msub><mml:mi>β</mml:mi><mml:mn>3</mml:mn></mml:msub></mml:mrow></mml:math></inline-formula></th><th valign="top" align="center" scope="col" rowspan="1" colspan="1"><inline-formula><mml:math id="me155"><mml:mrow><mml:msub><mml:mi>γ</mml:mi><mml:mn>2</mml:mn></mml:msub></mml:mrow></mml:math></inline-formula></th><th valign="top" align="center" scope="col" rowspan="1" colspan="1"><inline-formula><mml:math id="me156"><mml:mrow><mml:msub><mml:mi>γ</mml:mi><mml:mn>3</mml:mn></mml:msub></mml:mrow></mml:math></inline-formula></th><th valign="top" align="center" scope="col" rowspan="1" colspan="1"><inline-formula><mml:math id="me157"><mml:mrow><mml:msub><mml:mi>γ</mml:mi><mml:mn>4</mml:mn></mml:msub></mml:mrow></mml:math></inline-formula></th><th valign="top" align="center" scope="col" rowspan="1" colspan="1"><inline-formula><mml:math id="me158"><mml:mrow><mml:msub><mml:mi>γ</mml:mi><mml:mn>5</mml:mn></mml:msub></mml:mrow></mml:math></inline-formula></th></tr></thead><tbody><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">1</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.5108</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−1.9852</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.3863</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−3.7668</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.5652</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−1.6167</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−0.8285</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">2</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.5522</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.0165</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.4118</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−3.8006</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.6059</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−1.6544</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−0.8544</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">3</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.5129</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.0008</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.3791</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−3.7776</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.5840</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−1.6313</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−0.8338</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">4</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.5397</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.0142</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.4125</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−3.7951</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.5989</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−1.6518</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−0.8497</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">5</td><td valign="bottom" align="center" rowspan="1" colspan="1">−2.5202</td><td valign="bottom" align="center" rowspan="1" colspan="1">−1.9898</td><td valign="bottom" align="center" rowspan="1" colspan="1">−2.3855</td><td valign="bottom" align="center" rowspan="1" colspan="1">−3.7718</td><td valign="bottom" align="center" rowspan="1" colspan="1">−2.5675</td><td valign="bottom" align="center" rowspan="1" colspan="1">−1.6242</td><td valign="bottom" align="center" rowspan="1" colspan="1">−0.8322</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">6</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−3.4488</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.7384</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−3.2335</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−5.1665</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−3.4851</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.2005</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−1.1765</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">7</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−3.4497</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.7242</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−3.2277</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−5.1629</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−3.4768</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.2021</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−1.1821</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">8</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−3.4587</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.7354</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−3.2449</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−5.1674</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−3.4795</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.2056</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−1.1766</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">9</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−3.4402</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.7111</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−3.2167</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−5.1345</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−3.4641</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−2.1834</td><td valign="top" align="char" char="." rowspan="1" colspan="1">−1.1645</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1"></td><td colspan="3" valign="top" align="center" rowspan="1">SD Fixed Parameters</td><td colspan="4" valign="top" align="center" rowspan="1"> SD Threshold Parameters</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">1</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3362</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3242</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3381</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3525</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3479</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3310</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.2951</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">2</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3403</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3284</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3415</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3566</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3518</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3366</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3028</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">3</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3422</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3311</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3438</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3593</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3545</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3391</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3031</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">4</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3242</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3132</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3256</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3389</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3341</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3215</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.2903</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">5</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3321</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3210</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3338</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3479</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3429</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3289</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.2956</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">6</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4749</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4515</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4720</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.5324</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4923</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4490</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3910</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">7</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4801</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4569</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4770</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.5338</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4953</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4564</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.3997</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">8</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4791</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4571</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4756</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.5316</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4949</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4561</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4014</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">9</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4819</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4588</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4783</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.5352</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4981</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4576</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.4007</td></tr></tbody></table><table-wrap-foot><fn><p>SD, standard deviation.</p></fn></table-wrap-foot></table-wrap><fig id="fig2" fig-type="figure" position="float"><label>Figure 2</label><caption><p>Estimated probability of each category in the whole data set and of each location in model 9.</p></caption><graphic xlink:href="291f2"></graphic></fig><p>It is important to point out that the average probabilities of high infection (category 4) were very similar to the average probabilities of complete infection (category 5) for the whole data set and for each country, and that no infection (category 1) was the category with the lowest average probabilities (around 15%). These probability estimates were very different from estimates obtained based on raw frequencies (<xref ref-type="fig" rid="fig1">Figure 1</xref>) because they take into account the distribution of records across countries, the structure of lines, and the interaction between markers and environments (countries). The linear predictor used for doing this calculation was taken from model 9, which is given in <xref ref-type="disp-formula" rid="eq10">Equation (10)</xref>.</p><p>Estimates of variance components derived from the full data analysis are given in <xref ref-type="table" rid="t3">Table 3</xref>. The interaction (G×E) explained the largest proportion of the variance of liabilities, with estimated posterior means greater than 1.03 (models 6−9). The total variance explained by model 1 was 1.20, which is 1.7659 times smaller than the total variance captured by the most parameterized model (model 9), with a total variance of 2.1191. Also, model 2 (with E and G as main effects) only captured a total variance equal to 1.1911, which was 1.7791 times smaller than the total variance captured by model 9. The variance explained by models 3, 4, and 5 were, respectively, 1.7755, 1.7767, and 1.782 times smaller than the variance captured by model 9. However, models 6, 7, and 8 captured a total variance equal to 2.1321, 2.0967, and 2.116, respectively, which were almost identical to the total variance captured by model 9; model 6 captured the highest variability. It should be noted that in models 6, 7, 8, and 9, the percentages of variance explained by the interaction term (G×E) were 51.68, 49.42, 51.23, and 50.19%, respectively, clearly indicating that the interaction term was the most important source of variability in these models. It should be noted that if interactions were omitted, a large proportion of the variability in models 6 and 8 was not captured by main effects and the variance of the error term <inline-formula><mml:math id="me159"><mml:mrow><mml:msub><mml:mi>ε</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula> was the greatest source of variability. Another important finding was that the additive × additive epistatic terms explained 1.72, 1.06, 0.58, and 0.46% of the total variability in models 3, 5, 7, and 9, respectively, which means that adding this type of epistasis is not very important for this trait in these populations.</p><table-wrap id="t3" position="float"><label>Table 3</label><caption><title>Estimated variance components of the nine proposed models</title></caption><table frame="above" rules="groups"><col width="17.41%" span="1"></col><col width="14.05%" span="1"></col><col width="14.06%" span="1"></col><col width="14.06%" span="1"></col><col width="14.06%" span="1"></col><col width="14.06%" span="1"></col><col width="12.3%" span="1"></col><thead><tr><th valign="top" align="left" scope="col" rowspan="1" colspan="1">Model</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">L</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">G</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">G×G</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">G×E</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">G×G×E</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">TotVar</th></tr></thead><tbody><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">1</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.2000 (16.67)</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="char" char="." rowspan="1" colspan="1">1.2000</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">2</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.1911 (16.04)</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="char" char="." rowspan="1" colspan="1">1.1911</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">3</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.1730 (14.50)</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.0205 (1.72)</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="char" char="." rowspan="1" colspan="1">1.1935</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">4</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.0112 (0.94)</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.1815 (15.22)</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="char" char="." rowspan="1" colspan="1">1.1927</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">5</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.0090 (0.76)</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.1676 (14.09)</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.0126 (1.06)</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="char" char="." rowspan="1" colspan="1">1.1892</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">6</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.0303 (1.42)</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="char" char="." rowspan="1" colspan="1">1.1018 (51.68)</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="char" char="." rowspan="1" colspan="1">2.1321</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">7</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.0165 (0.79)</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.0121 (0.58)</td><td valign="top" align="char" char="." rowspan="1" colspan="1">1.0362 (49.42)</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.0319 (1.52)</td><td valign="top" align="char" char="." rowspan="1" colspan="1">2.0967</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">8</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.0117 (0.55)</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.0202 (0.95)</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="char" char="." rowspan="1" colspan="1">1.0841 (51.23)</td><td valign="top" align="left" rowspan="1" colspan="1"></td><td valign="top" align="char" char="." rowspan="1" colspan="1">2.116</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">9</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.0080 (0.38)</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.0122 (0.58)</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.0097 (0.46)</td><td valign="top" align="char" char="." rowspan="1" colspan="1">1.0636 (50.19)</td><td valign="top" align="char" char="." rowspan="1" colspan="1">0.0256 (1.21)</td><td valign="top" align="char" char="." rowspan="1" colspan="1">2.1191</td></tr></tbody></table><table-wrap-foot><fn><p>Numbers in parenthesis are the percentages of variance explained by each component. L, line; G, marker covariates; G×G, additive × additive epistasis term; G×E, environment × marker interaction; G×G×E, additive × additive epistasis × environment interaction term; TotVar, total variance explained by each model including the variance of <inline-formula><mml:math id="me160"><mml:mrow><mml:msub><mml:mi>ε</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi><mml:mi>k</mml:mi></mml:mrow></mml:msub><mml:mo>,</mml:mo></mml:mrow></mml:math></inline-formula> which is equal to 1.</p></fn></table-wrap-foot></table-wrap><p><xref ref-type="table" rid="t4">Table 4</xref> presents the Brier scores of the validation samples for each country and for the nine models. Because phenotype was ordinal categorical, we used Brier scores instead of Pearson’s correlation coefficient for assessing prediction accuracy, because the Brier scores are bounded between 0 and 1, and values closer to zero imply better prediction accuracy. In <xref ref-type="table" rid="t4">Table 4</xref>, models 6, 7, 8, and 9 presented the lowest Brier scores, so these four models had better prediction ability. Models 1−5 showed the worst prediction ability. In Colombia, the prediction ability of model 9 was 19.85% greater than that of model 1. In Mexico, the increase in the prediction ability of model 9 over model 1 (the simplest model) was 11.26%, while in Zimbabwe this increase was 8.71%. Although <xref ref-type="table" rid="t3">Table 3</xref> showed that inclusion of the interaction term (G×E) explained the largest proportion of variability in models 6 and 8, this did not produce a large increase in the prediction ability of these two models over that of the models without interactions (models 1, 2, and 4). This may be due to three reasons: (1) predictive power differs from goodness of fit, that is, a model may fit a particular data set well even though the predictive power the model provides is small; (2) even though the added interaction term explained the largest proportion of variability in models 6 and 8, the total variability explained in all models was not large; and (3) the set of lines was relatively diverse. However, including the interaction terms G×E plus the additive × additive epistatic terms [main (<inline-formula><mml:math id="me161"><mml:mrow><mml:msub><mml:mi>g</mml:mi><mml:mrow><mml:mi>A</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:math></inline-formula> and interaction term (<inline-formula><mml:math id="me162"><mml:mrow><mml:msub><mml:mi>g</mml:mi><mml:mi>A</mml:mi></mml:msub><mml:msub><mml:mi>E</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mi>j</mml:mi></mml:mrow></mml:msub><mml:mo stretchy="false">)</mml:mo><mml:mo stretchy="false">]</mml:mo></mml:mrow></mml:math></inline-formula> did not improve predictions by much, and the inclusion of these terms did not explain a large percentage of the total variability.</p><table-wrap id="t4" position="float"><label>Table 4</label><caption><title>Brier scores (mean, minimum and maximum; smaller indicates better prediction) evaluated for the validation samples</title></caption><table frame="hsides" rules="groups"><col width="12.61%" span="1"></col><col width="10.06%" span="1"></col><col width="10.05%" span="1"></col><col width="10.05%" span="1"></col><col width="10.32%" span="1"></col><col width="9.13%" span="1"></col><col width="9.49%" span="1"></col><col width="9.13%" span="1"></col><col width="9.13%" span="1"></col><col width="10.03%" span="1"></col><thead><tr><th rowspan="2" valign="top" align="left" scope="col" colspan="1">Model</th><th colspan="3" valign="top" align="center" scope="colgroup" rowspan="1">Colombia</th><th colspan="3" valign="top" align="center" scope="colgroup" rowspan="1">Zimbabwe</th><th colspan="3" valign="top" align="center" scope="colgroup" rowspan="1">Mexico</th></tr><tr><th valign="top" colspan="1" align="center" scope="colgroup" rowspan="1">Mean</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">Min</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">Max</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">Mean</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">Min</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">Max</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">Mean</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">Min</th><th valign="top" align="center" scope="col" rowspan="1" colspan="1">Max</th></tr></thead><tbody><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">1</td><td valign="top" align="center" rowspan="1" colspan="1">0.3924</td><td valign="top" align="center" rowspan="1" colspan="1">0.3798</td><td valign="top" align="center" rowspan="1" colspan="1">0.4115</td><td valign="top" align="center" rowspan="1" colspan="1">0.3617</td><td valign="top" align="center" rowspan="1" colspan="1">0.3554</td><td valign="top" align="center" rowspan="1" colspan="1">0.3698</td><td valign="top" align="center" rowspan="1" colspan="1">0.3507</td><td valign="top" align="center" rowspan="1" colspan="1">0.3386</td><td valign="top" align="center" rowspan="1" colspan="1">0.3604</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">2</td><td valign="top" align="center" rowspan="1" colspan="1">0.3869</td><td valign="top" align="center" rowspan="1" colspan="1">0.3744</td><td valign="top" align="center" rowspan="1" colspan="1">0.4011</td><td valign="top" align="center" rowspan="1" colspan="1">0.3611</td><td valign="top" align="center" rowspan="1" colspan="1">0.3542</td><td valign="top" align="center" rowspan="1" colspan="1">0.3663</td><td valign="top" align="center" rowspan="1" colspan="1">0.3434</td><td valign="top" align="center" rowspan="1" colspan="1">0.3331</td><td valign="top" align="center" rowspan="1" colspan="1">0.3572</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">3</td><td valign="top" align="center" rowspan="1" colspan="1">0.3845</td><td valign="top" align="center" rowspan="1" colspan="1">0.3733</td><td valign="top" align="center" rowspan="1" colspan="1">0.4021</td><td valign="top" align="center" rowspan="1" colspan="1">0.3628</td><td valign="top" align="center" rowspan="1" colspan="1">0.3559</td><td valign="top" align="center" rowspan="1" colspan="1">0.3701</td><td valign="top" align="center" rowspan="1" colspan="1">0.3433</td><td valign="top" align="center" rowspan="1" colspan="1">0.3302</td><td valign="top" align="center" rowspan="1" colspan="1">0.3591</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">4</td><td valign="top" align="center" rowspan="1" colspan="1">0.3856</td><td valign="top" align="center" rowspan="1" colspan="1">0.3706</td><td valign="top" align="center" rowspan="1" colspan="1">0.4024</td><td valign="top" align="center" rowspan="1" colspan="1">0.3621</td><td valign="top" align="center" rowspan="1" colspan="1">0.3538</td><td valign="top" align="center" rowspan="1" colspan="1">0.3697</td><td valign="top" align="center" rowspan="1" colspan="1">0.3431</td><td valign="top" align="center" rowspan="1" colspan="1">0.3337</td><td valign="top" align="center" rowspan="1" colspan="1">0.3526</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">5</td><td valign="top" align="center" rowspan="1" colspan="1">0.3860</td><td valign="top" align="center" rowspan="1" colspan="1">0.3734</td><td valign="top" align="center" rowspan="1" colspan="1">0.4012</td><td valign="top" align="center" rowspan="1" colspan="1">0.3619</td><td valign="top" align="center" rowspan="1" colspan="1">0.3528</td><td valign="top" align="center" rowspan="1" colspan="1">0.3734</td><td valign="top" align="center" rowspan="1" colspan="1">0.3448</td><td valign="top" align="center" rowspan="1" colspan="1">0.3251</td><td valign="top" align="center" rowspan="1" colspan="1">0.3598</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">6</td><td valign="top" align="center" rowspan="1" colspan="1">0.3261</td><td valign="top" align="center" rowspan="1" colspan="1">0.3121</td><td valign="top" align="center" rowspan="1" colspan="1">0.3402</td><td valign="top" align="center" rowspan="1" colspan="1">0.3337</td><td valign="top" align="center" rowspan="1" colspan="1">0.3249</td><td valign="top" align="center" rowspan="1" colspan="1">0.3413</td><td valign="top" align="center" rowspan="1" colspan="1">0.3145</td><td valign="top" align="center" rowspan="1" colspan="1">0.2972</td><td valign="top" align="center" rowspan="1" colspan="1">0.3295</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">7</td><td valign="top" align="center" rowspan="1" colspan="1">0.3315</td><td valign="top" align="center" rowspan="1" colspan="1">0.3170</td><td valign="top" align="center" rowspan="1" colspan="1">0.3427</td><td valign="top" align="center" rowspan="1" colspan="1">0.3308</td><td valign="top" align="center" rowspan="1" colspan="1">0.3214</td><td valign="top" align="center" rowspan="1" colspan="1">0.3363</td><td valign="top" align="center" rowspan="1" colspan="1">0.3183</td><td valign="top" align="center" rowspan="1" colspan="1">0.3003</td><td valign="top" align="center" rowspan="1" colspan="1">0.3364</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">8</td><td valign="top" align="center" rowspan="1" colspan="1">0.3249</td><td valign="top" align="center" rowspan="1" colspan="1">0.3141</td><td valign="top" align="center" rowspan="1" colspan="1">0.3417</td><td valign="top" align="center" rowspan="1" colspan="1">0.3345</td><td valign="top" align="center" rowspan="1" colspan="1">0.3247</td><td valign="top" align="center" rowspan="1" colspan="1">0.3441</td><td valign="top" align="center" rowspan="1" colspan="1">0.3189</td><td valign="top" align="center" rowspan="1" colspan="1">0.3094</td><td valign="top" align="center" rowspan="1" colspan="1">0.3277</td></tr><tr><td valign="top" align="left" scope="row" rowspan="1" colspan="1">9</td><td valign="top" align="center" rowspan="1" colspan="1">0.3274</td><td valign="top" align="center" rowspan="1" colspan="1">0.3159</td><td valign="top" align="center" rowspan="1" colspan="1">0.3401</td><td valign="top" align="center" rowspan="1" colspan="1">0.3327</td><td valign="top" align="center" rowspan="1" colspan="1">0.3155</td><td valign="top" align="center" rowspan="1" colspan="1">0.3455</td><td valign="top" align="center" rowspan="1" colspan="1">0.3152</td><td valign="top" align="center" rowspan="1" colspan="1">0.2981</td><td valign="top" align="center" rowspan="1" colspan="1">0.3280</td></tr></tbody></table></table-wrap></sec><sec sec-type="discussion" id="s24"><title>Discussion</title><p>GS offers important opportunities to improve genetic gains in plant and animal breeding programs; however, more powerful estimation and prediction methods are required to fully exploit the advantages of GS. Most GS methods assume a Gaussian response; however, many important traits in plant breeding, such as disease resistance, percent protein content, and proportion of seed or plant damage, are not normally distributed and need special treatment. In this study, we extended the GBLUP method to ordinal categorical responses. To this end, we used the implementation available in BGLR with a probit link that can accommodate both binary and ordinal traits (<xref rid="bib25" ref-type="bibr">Pérez-Rodríguez and de los Campos 2014</xref>). To our knowledge, this is the first study in GS that uses an ordinal threshold model to analyze a categorical response in plant breeding.</p><p>Most non-normal response variables are not linear with respect to model parameters and, unlike normal responses, the variance of the phenotypes is dependent on the mean. The GLMM uses: (1) a link function (typically nonlinear on the parameters) to specify the relationship between the mean of the response variable and a linear model; and (2) a variance function to describe the relationship between the mean and the variance of the distribution of the response variable, which enables it to relate traditional linear regression to non-normal data. This gives the GLMM framework a stronger basis for hypothesis testing and for more precise estimates of fixed and random effects. This statement applies to response variables coming from many different distributions, and not only to ordinal categorical responses.</p><p>GS methods have been implemented for threshold (ordinal categorical) traits in animal breeding. For example, <xref rid="bib16" ref-type="bibr">González-Recio and Forni (2011)</xref> developed versions of BayesA, Bayesian Lasso, and two machine learning methods for dichotomous traits; <xref rid="bib31" ref-type="bibr">Villanueva <italic>et al.</italic> (2011)</xref> also developed a version of BayesB for dichotomous traits; <xref rid="bib32" ref-type="bibr">Wang <italic>et al.</italic> (2012)</xref> implemented methods BayesA, BayesB, and BayesC<inline-formula><mml:math id="me163"><mml:mi>π</mml:mi></mml:math></inline-formula> for ordinal categorical traits; and <xref rid="bib21" ref-type="bibr">Kizilkaya <italic>et al.</italic> (2014)</xref> implemented BayesC<inline-formula><mml:math id="me164"><mml:mrow><mml:mi>π</mml:mi></mml:mrow></mml:math></inline-formula> for ordinal categorical traits. All these studies concluded that the traditional GS linear models are not suitable for threshold traits, since the basic assumptions of the linear model are violated, which produces a considerable reduction in prediction accuracy. However, to our knowledge, no plant breeding study so far has addressed the analysis of categorical traits. Our study fills this gap by providing a full description of a multithreshold GBLUP model (TGBLUP) and an empirical evaluation based on real data.</p><p>Prediction accuracy in GS usually is assessed using the sample correlation between predictions and phenotypes. However, this metric is not appropriate for categorical outcomes. For this reason, we used the Brier score for assessing the prediction ability of the proposed threshold model, which assigns a numerical score based on the predictive distribution. This scoring rule is “strictly proper” in the sense that it maximizes the expected score of an observation (and is unique) and allows making fair comparisons among models. To our knowledge, our study is the first one in GS to use this score for assessing prediction accuracy.</p><p>We presented nine specifications of the TGBLUP model which differ on the type of effects included. We considered main effects models and models for interactions between genetic factors (<italic>e.g.</italic>, additive ×additive epistatic) and between genetic and environmental factors <italic>e.g.</italic>, additive ×additive× environment interaction. For specifying these interactions, we used the framework proposed by <xref rid="bib20" ref-type="bibr">Jarquín <italic>et al.</italic> (2014)</xref> within a threshold model framework. We found that models that take into account interactions (models 6−9) explained almost two times more variability than those without interaction (models 1−5). However, the interaction that was clearly most important was the G×E term; this effect captured the largest proportion of the total variability explained by the models (51.68% in model 6 and 50.19% in model 9), and was the one that led to the largest gain in prediction accuracy. This result is in agreement with previous studies that have highlighted the importance of modeling G×E in GS in plant breeding (<xref rid="bib3" ref-type="bibr">Burgueño <italic>et al.</italic> 2012</xref>; <xref rid="bib20" ref-type="bibr">Jarquin <italic>et al.</italic> 2014</xref>).</p><p>Inclusion of additive × additive epistatic terms did not explain much of the total variability, and did not help much to improve prediction ability. Also, we found that estimated threshold parameters could be clustered into two groups, one for those without interactions (models 1−5), and another for those with interaction terms (models 6−9). This was because including interactions increases the variance of the liability score and, therefore, changes in threshold values are needed to accommodate the observed probabilities of each of the categories.</p><p>Finally, although the analyses presented here used Gaussian priors for marker effects, the multi-threshold model used in this study can be implemented with any of the priors commonly used in GS, including those that induce differential shrinkage of estimates of effects or a combination of variable selection and shrinkage. The current implementation of BGLR allows users to do this.</p><p>We extended the GBLUP, a model commonly used in GS for analyses of normal traits, to situations where the response is ordinal (TGBLUP). We provided a detailed description of the model and introduced a metric, the Brier score, for assessing prediction accuracy of ordinal categorical outcomes. We presented an empirical evaluation using a (real) data set of 278 maize lines genotyped with 46,347 SNPs and evaluated for disease resistance using an ordinal categorical scoring system in three environments (Colombia, Zimbabwe, and Mexico). A total of nine models were used. In addition, we provide details of the R code used to implement these models using the BGLR package.</p><p>Our results highlight the importance of including G×E (capturing at least 49.42% of the total variability); when this interaction was taken into account, this increased the total variability explained by these models and increased prediction accuracy between 8 and 19% relative to models based on main effects only. Considering additive × additive epistasis did not produce a sizable increase in prediction accuracy.</p></sec></body><back><fn-group><fn id="fn1"><p><italic>Communicating editor: D. J. de Koning</italic></p></fn></fn-group><app-group><app id="app1"><title>Appendix 1</title><sec id="s25"><title>FITTING MODEL 9 TO RANDOMLY SAMPLED TRAINING AND TESTING SETS</title><p>library(BGLR)</p><p>library(matrixcalc)</p><p># X, data frame containing the elements described below;</p><p># - X$loc: (n×1), a factor giving the IDs for the environments (location);</p><p># - X$line: (n×1), a factor giving the IDs for the varieties;</p><p># - X$y: (n×1), a numeric vector for rating the 5 levels of the disease;</p><p># - M: a matrix containing the genetic information (dimensions equal to the number of lines by the</p><p># - number of SNPs [46,347]).</p><p>### Note: To replicate the results presented in this article, the sampling of training-testing sets must be done #according to the methods described in the article (70% training and 30% testing); this example simply #illustrates how to fit model 9 for a randomly chosen testing data set. The code for the other model is similar.</p><p>#Calculating the marker-derived genomic relationship matrix (GRM)</p><p>M<-scale(M,center=TRUE,scale=TRUE)</p><p>G<-tcrossprod(M)/ncol(M)</p><p># Incidence matrix and covariance for main effects of environments.</p><p>ZE<-model.matrix(∼factor(X$loc)-1) </p><p>KE=tcrossprod(ZE);</p><p># Incidence matrix and covariance for main effects of lines.</p><p>X$lines<-factor(x=X$lines, levels=rownames(M), ordered=TRUE)</p><p>ZL<-model.matrix(∼X$lines-1)</p><p>KL=tcrossprod(ZL);</p><p># Genetic covariance structure of genetic effects in the full data</p><p>KG= ZL%*%G%*%t(ZL);</p><p># Epistasis additive × additive covariance structure in the full data</p><p>GA=hadamard.prod(G,G)</p><p>KGG= ZL%*%GA%*%t(ZL);</p><p># G×E covariance structure for the full data set</p><p>KGE=hadamard.prod(KG, KE)</p><p>diag(KGE)=diag(KGE)</p><p>KGE=KGE/mean(diag(KGE))</p><p># GG×E Epistasis additive × additive structure for the full data set</p><p>KGGE=hadamard.prod(KGG, KE)</p><p>diag(KGGE)=diag(KGGE)</p><p>KGGE=KGGE/mean(diag(KGGE))</p><p>yNA<- X$y; n=length(X$y); p=round(0.30*n);</p><p>tst<-sample(1:nrow(X), size=p, replace=FALSE)</p><p>yNA[tst]<-NA</p><p>ETA<-list(ENV=list(K=ZE, model=’FIXED’), LINE=list(K=KL, model=’RKHS’),</p><p>G=list(K=KG, model=’RKHS’), GG=list(K=KGG, model=’RKHS’), </p><p>GE=list(K=KGE, model=’RKHS’), GGE=list(K=KGGE, model=’RKHS’))</p><p>fm9<-BGLR(y=yNA,response_type=’ordinal’, saveAt=’M9_’, nIter=52000, burnIn=6000)</p></sec></app></app-group><ref-list><title>Literature Cited</title><ref id="bib1"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Atkinson</surname><given-names>L.</given-names></name></person-group>, <year>1988</year> <article-title>The measurement-statistics controversy: factor analysis and subinterval data.</article-title><source>Bull. Psychon. Soc.</source><volume>26</volume>: <fpage>361</fpage>–<lpage>364</lpage>.</mixed-citation></ref><ref id="bib2"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Brier</surname><given-names>G. W.</given-names></name></person-group>, <year>1950</year> <article-title>Verification of forecasts expressed in terms of probability.</article-title><source>Mon. Weather Rev.</source><volume>78</volume>: <fpage>1</fpage>–<lpage>3</lpage>.</mixed-citation></ref><ref id="bib3"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Burgueño</surname><given-names>J.</given-names></name><name><surname>de los Campos</surname><given-names>G.</given-names></name><name><surname>Weigel</surname><given-names>K.</given-names></name><name><surname>Crossa</surname><given-names>J.</given-names></name></person-group>, <year>2012</year> <article-title>Genomic prediction of breeding values when modeling genotype×environment interaction using pedigree and dense molecular markers.</article-title><source>Crop Sci.</source><volume>52</volume>: <fpage>707</fpage>.</mixed-citation></ref><ref id="bib4"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Crossa</surname><given-names>J.</given-names></name><name><surname>de los Campos</surname><given-names>G.</given-names></name><name><surname>Pérez-Rodríguez</surname><given-names>P.</given-names></name><name><surname>Gianola</surname><given-names>D.</given-names></name><name><surname>Burgueño</surname><given-names>J.</given-names></name><etal></etal></person-group><year>2010</year> <article-title>Prediction of genetic values of quantitative traits in plant breeding using pedigree and molecular markers.</article-title><source>Genetics</source><volume>186</volume>: <fpage>713</fpage>–<lpage>724</lpage>.<pub-id pub-id-type="pmid">20813882</pub-id></mixed-citation></ref><ref id="bib5"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Crossa</surname><given-names>J.</given-names></name><name><surname>Pérez-Rodríguez</surname><given-names>P.</given-names></name><name><surname>de los Campos</surname><given-names>G.</given-names></name><name><surname>Mahuku</surname><given-names>G.</given-names></name><name><surname>Dreisigacker</surname><given-names>S.</given-names></name><etal></etal></person-group><year>2011</year> <article-title>Genomic selection and prediction in plant breeding.</article-title><source>J. Crop Improv.</source><volume>25</volume>: <fpage>239</fpage>–<lpage>261</lpage>.</mixed-citation></ref><ref id="bib6"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Crossa</surname><given-names>J.</given-names></name><name><surname>Beyene</surname><given-names>Y.</given-names></name><name><surname>Kassa</surname><given-names>S.</given-names></name><name><surname>Pérez-Rodríguez</surname><given-names>P.</given-names></name><name><surname>Hickey</surname><given-names>J. M.</given-names></name><etal></etal></person-group><year>2013</year><comment>a</comment> <article-title>Genomic prediction in maize breeding populations with genotyping-by-sequencing.</article-title><source>G3 (Bethesda)</source><volume>3</volume>: <fpage>1903</fpage>–<lpage>1926</lpage>.<pub-id pub-id-type="pmid">24022750</pub-id></mixed-citation></ref><ref id="bib7"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Crossa</surname><given-names>J.</given-names></name><name><surname>Pérez-Rodríguez</surname><given-names>P.</given-names></name><name><surname>Hickey</surname><given-names>J.</given-names></name><name><surname>Burgueño</surname><given-names>J.</given-names></name><name><surname>Ornella</surname><given-names>L.</given-names></name><etal></etal></person-group><year>2013</year><comment>b</comment> <article-title>Genomic prediction in CIMMYT maize and wheat breeding programs.</article-title><source>Heredity</source><volume>112</volume>: <fpage>48</fpage>–<lpage>60</lpage>.<pub-id pub-id-type="pmid">23572121</pub-id></mixed-citation></ref><ref id="bib8"><mixed-citation publication-type="webpage">de los Campos, G., and P. Pérez-Rodríguez, 2013 BGLR: Bayesian Generalized Linear Regression. R package version 1.0.3. Availabe at: <ext-link ext-link-type="uri" xlink:href="http://CRAN.R-project.org/package=BGLR">http://CRAN.R-project.org/package=BGLR</ext-link>. Accessed December 30, 2014.</mixed-citation></ref><ref id="bib9"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>de los Campos</surname><given-names>G.</given-names></name><name><surname>Naya</surname><given-names>H.</given-names></name><name><surname>Gianola</surname><given-names>D.</given-names></name><name><surname>Crossa</surname><given-names>J.</given-names></name><name><surname>Legarra</surname><given-names>A.</given-names></name><etal></etal></person-group><year>2009</year> <article-title>Predicting quantitative traits with regression models for dense molecular markers and pedigree.</article-title><source>Genetics</source><volume>182</volume>: <fpage>375</fpage>–<lpage>385</lpage>.<pub-id pub-id-type="pmid">19293140</pub-id></mixed-citation></ref><ref id="bib10"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>de los Campos</surname><given-names>G.</given-names></name><name><surname>Gianola</surname><given-names>D.</given-names></name><name><surname>Rosa</surname><given-names>G. J. M.</given-names></name><name><surname>Weigel</surname><given-names>K.</given-names></name><name><surname>Crossa</surname><given-names>J.</given-names></name></person-group>, <year>2010</year> <article-title>Semi-parametric genomic-enabled prediction of genetic values using reproducing kernel Hilbert spaces methods.</article-title><source>Genet. Res.</source><volume>92</volume>: <fpage>295</fpage>–<lpage>308</lpage>.</mixed-citation></ref><ref id="bib11"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Garthwaite</surname><given-names>P. H.</given-names></name><name><surname>Kadane</surname><given-names>J. B.</given-names></name><name><surname>O’Hagan</surname><given-names>A.</given-names></name></person-group>, <year>2005</year> <article-title>Statistical methods for eliciting probability distributions.</article-title><source>J. Am. Stat. Assoc.</source><volume>100</volume>: <fpage>680</fpage>–<lpage>701</lpage>.</mixed-citation></ref><ref id="bib12"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Gianola</surname><given-names>D.</given-names></name></person-group>, <year>1980</year> <article-title>A method of sire evaluation for dichotomies.</article-title><source>J. Anim. Sci.</source><volume>51</volume>: <fpage>1266</fpage>–<lpage>1271</lpage>.<pub-id pub-id-type="pmid">7204270</pub-id></mixed-citation></ref><ref id="bib13"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Gianola</surname><given-names>D.</given-names></name></person-group>, <year>1982</year> <article-title>Theory and analysis of threshold characters.</article-title><source>J. Anim. Sci.</source><volume>54</volume>: <fpage>1079</fpage>–<lpage>1096</lpage>.</mixed-citation></ref><ref id="bib14"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Gianola</surname><given-names>D.</given-names></name><name><surname>Foulley</surname><given-names>J. L.</given-names></name></person-group>, <year>1983</year> <article-title>Sire evaluation for ordered categorical data with a threshold model.</article-title><source>Genet. Sel. Evol.</source><volume>15</volume>: <fpage>1</fpage>–<lpage>23</lpage>.<pub-id pub-id-type="pmid">22879102</pub-id></mixed-citation></ref><ref id="bib15"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>González-Camacho</surname><given-names>J. M.</given-names></name><name><surname>de los Campos</surname><given-names>G.</given-names></name><name><surname>Pérez-Rodríguez</surname><given-names>P.</given-names></name><name><surname>Gianola</surname><given-names>D.</given-names></name><name><surname>Cairns</surname><given-names>J. E.</given-names></name><etal></etal></person-group><year>2012</year> <article-title>Genome-enabled prediction of genetic values using radial basis function neural networks.</article-title><source>Theor. Appl. Genet.</source><volume>125</volume>: <fpage>759</fpage>–<lpage>771</lpage>.<pub-id pub-id-type="pmid">22566067</pub-id></mixed-citation></ref><ref id="bib16"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>González-Recio</surname><given-names>O.</given-names></name><name><surname>Forni</surname><given-names>S.</given-names></name></person-group>, <year>2011</year> <article-title>Genome-wide prediction of discrete traits using Bayesian regressions and machine learning.</article-title><source>Genet. Sel. Evol.</source><volume>43</volume>: <fpage>7</fpage>.<pub-id pub-id-type="pmid">21329522</pub-id></mixed-citation></ref><ref id="bib17"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Heffner</surname><given-names>E. L.</given-names></name><name><surname>Sorrells</surname><given-names>M. E.</given-names></name><name><surname>Jannink</surname><given-names>J.-L.</given-names></name></person-group>, <year>2009</year> <article-title>Genomic selection for crop improvement.</article-title><source>Crop Sci.</source><volume>49</volume>: <fpage>1</fpage>–<lpage>12</lpage>.</mixed-citation></ref><ref id="bib18"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Heffner</surname><given-names>E. L.</given-names></name><name><surname>Lorenz</surname><given-names>A. J.</given-names></name><name><surname>Jannink</surname><given-names>J.-L.</given-names></name><name><surname>Sorrells</surname><given-names>M. E.</given-names></name></person-group>, <year>2010</year> <article-title>Plant breeding with genomic selection: gain per unit time and cost.</article-title><source>Crop Sci.</source><volume>50</volume>: <fpage>1681</fpage>–<lpage>1690</lpage>.</mixed-citation></ref><ref id="bib19"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Heslot</surname><given-names>N.</given-names></name><name><surname>Yang</surname><given-names>H.-P.</given-names></name><name><surname>Sorrells</surname><given-names>M. E.</given-names></name><name><surname>Jannink</surname><given-names>J.-L.</given-names></name></person-group>, <year>2012</year> <article-title>Genomic selection in plant breeding: a comparison of models.</article-title><source>Crop Sci.</source><volume>52</volume>: <fpage>146</fpage>.</mixed-citation></ref><ref id="bib20"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Jarquín</surname><given-names>D.</given-names></name><name><surname>Crossa</surname><given-names>J.</given-names></name><name><surname>Lacaze</surname><given-names>X.</given-names></name><name><surname>Cheyron</surname><given-names>P.</given-names></name><name><surname>Daucourt</surname><given-names>J.</given-names></name><etal></etal></person-group><year>2014</year> <article-title>A reaction norm model for genomic selection using high-dimensional genomic and environmental data.</article-title><source>Theor. Appl. Genet.</source><volume>127</volume>: <fpage>595</fpage>–<lpage>607</lpage>.<pub-id pub-id-type="pmid">24337101</pub-id></mixed-citation></ref><ref id="bib21"><mixed-citation publication-type="confproc"><person-group person-group-type="author"><name><surname>Kizilkaya</surname><given-names>K.</given-names></name><name><surname>Tait</surname><given-names>R. G.</given-names></name><name><surname>Garrick</surname><given-names>D. J.</given-names></name><name><surname>Fernando</surname><given-names>R. L.</given-names></name><name><surname>Reecy</surname><given-names>J. M.</given-names></name></person-group>, <year>2014</year> <article-title>Whole genome analysis of infectious bovine kerato conjunctivitis in Angus cattle using Bayesian threshold models.</article-title><source>BMC Proc.</source><volume>5</volume>: <fpage>S22</fpage>.<pub-id pub-id-type="pmid">21645302</pub-id></mixed-citation></ref><ref id="bib22"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Kneib</surname><given-names>T.</given-names></name><name><surname>Baumgartner</surname><given-names>B.</given-names></name><name><surname>Steiner</surname><given-names>W. J.</given-names></name></person-group>, <year>2007</year> <article-title>Semiparametric multinomial logit models for analysing consumer choice behaviour.</article-title><source>AStA Adv. Stat. Anal.</source><volume>91</volume>: <fpage>225</fpage>–<lpage>244</lpage>.</mixed-citation></ref><ref id="bib23"><mixed-citation publication-type="book"><person-group person-group-type="author"><name><surname>Littell</surname><given-names>R. C.</given-names></name><name><surname>Stroup</surname><given-names>W.</given-names></name><name><surname>Freund</surname><given-names>R. J.</given-names></name></person-group>, <year>2002</year> <source>SAS for Linear Models</source>. <edition>ed. 4</edition><publisher-name>SAS Institute</publisher-name>, <publisher-loc>Cary, NC</publisher-loc>.</mixed-citation></ref><ref id="bib24"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Meuwissen</surname><given-names>T. H. E.</given-names></name><name><surname>Hayes</surname><given-names>B. J.</given-names></name><name><surname>Goddard</surname><given-names>M. E.</given-names></name></person-group>, <year>2001</year> <article-title>Prediction of total genetic values using genome-wide dense marker maps.</article-title><source>Genetics</source><volume>157</volume>: <fpage>1819</fpage>–<lpage>1829</lpage>.<pub-id pub-id-type="pmid">11290733</pub-id></mixed-citation></ref><ref id="bib25"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Pérez-Rodríguez</surname><given-names>P.</given-names></name><name><surname>de los Campos</surname><given-names>G.</given-names></name></person-group>, <year>2014</year> <article-title>Genome-wide regression & prediction with the BGLR statistical package.</article-title><source>Genetics</source><volume>198</volume>: <fpage>483</fpage>–<lpage>495</lpage>.<pub-id pub-id-type="pmid">25009151</pub-id></mixed-citation></ref><ref id="bib26"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Pérez-Rodríguez</surname><given-names>P.</given-names></name><name><surname>de los Campos</surname><given-names>G.</given-names></name><name><surname>Crossa</surname><given-names>J.</given-names></name><name><surname>Gianola</surname><given-names>D.</given-names></name></person-group>, <year>2010</year> <article-title>Genomic-enabled prediction based on molecular markers and pedigree using the Bayesian Linear Regression Package in R.</article-title><source>Plant Genome</source><volume>3</volume>: <fpage>106</fpage>–<lpage>116</lpage>.<pub-id pub-id-type="pmid">21566722</pub-id></mixed-citation></ref><ref id="bib27"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Pérez-Rodríguez</surname><given-names>P.</given-names></name><name><surname>Gianola</surname><given-names>D.</given-names></name><name><surname>González-Camacho</surname><given-names>J. M.</given-names></name><name><surname>Crossa</surname><given-names>J.</given-names></name><name><surname>Manes</surname><given-names>Y.</given-names></name><name><surname>Dreisigacker,</surname><given-names>S.</given-names></name></person-group><year>2012</year> <article-title>Comparison between linear and non-parametric models for genome-enabled prediction in wheat.</article-title><source>G3 (Bethesda)</source><volume>2</volume>: <fpage>1595</fpage>–<lpage>1605</lpage>.<pub-id pub-id-type="pmid">23275882</pub-id></mixed-citation></ref><ref id="bib28"><mixed-citation publication-type="webpage">R Core Team, 2014 R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Available at: <ext-link ext-link-type="uri" xlink:href="http://www.R-project.org/">http://www.R-project.org/</ext-link>. Accessed December 30, 2014.</mixed-citation></ref><ref id="bib29"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Sorensen</surname><given-names>D. A.</given-names></name><name><surname>Andersen</surname><given-names>S.</given-names></name><name><surname>Gianola</surname><given-names>D.</given-names></name><name><surname>Korsgaard</surname><given-names>I.</given-names></name></person-group>, <year>1995</year> <article-title>Bayesian inference in threshold models using Gibbs sampling.</article-title><source>Genet. Sel. Evol.</source><volume>27</volume>: <fpage>229</fpage>–<lpage>249</lpage>.</mixed-citation></ref><ref id="bib30"><mixed-citation publication-type="book"><person-group person-group-type="author"><name><surname>Stroup</surname><given-names>W.</given-names></name></person-group>, <year>2012</year> <source><italic>Generalized Linear Mixed Models: Modern Concepts</italic>, <italic>Methods and Applications</italic></source>. <publisher-name>CRC Press</publisher-name>, <publisher-loc>Boca Raton, FL</publisher-loc>.</mixed-citation></ref><ref id="bib31"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Villanueva</surname><given-names>B.</given-names></name><name><surname>Fernandez</surname><given-names>J.</given-names></name><name><surname>Garcia-Cortes</surname><given-names>L. A.</given-names></name><name><surname>Varona</surname><given-names>L.</given-names></name><name><surname>Daetwyler</surname><given-names>H. D.</given-names></name><etal></etal></person-group><year>2011</year> <article-title>Accuracy of genome-wide evaluation for disease resistance in aquaculture breeding programs.</article-title><source>J. Anim. Sci.</source><volume>89</volume>: <fpage>3433</fpage>–<lpage>3442</lpage>.<pub-id pub-id-type="pmid">21742941</pub-id></mixed-citation></ref><ref id="bib32"><mixed-citation publication-type="journal"><person-group person-group-type="author"><name><surname>Wang</surname><given-names>C. L.</given-names></name><name><surname>Ding</surname><given-names>X. D.</given-names></name><name><surname>Wang</surname><given-names>J. Y.</given-names></name><name><surname>Liu</surname><given-names>J. F.</given-names></name><name><surname>Fu</surname><given-names>W. X.</given-names></name><etal></etal></person-group><year>2012</year> <article-title>Bayesian methods for estimating GEBVs of threshold traits.</article-title><source>Heredity</source><volume>110</volume>: <fpage>213</fpage>–<lpage>219</lpage>.<pub-id pub-id-type="pmid">23149458</pub-id></mixed-citation></ref></ref-list></back></pmc></record>Pour manipuler ce document sous Unix (Dilib)
EXPLOR_STEP=$WICRI_ROOT/Ticri/CIDE/explor/TelematiV1/Data/Pmc/Corpus
HfdSelect -h $EXPLOR_STEP/biblio.hfd -nk 000452 | SxmlIndent | more
Ou
HfdSelect -h $EXPLOR_AREA/Data/Pmc/Corpus/biblio.hfd -nk 000452 | SxmlIndent | more
Pour mettre un lien sur cette page dans le réseau Wicri
{{Explor lien
|wiki= Ticri/CIDE
|area= TelematiV1
|flux= Pmc
|étape= Corpus
|type= RBID
|clé=
|texte=
}}
| This area was generated with Dilib version V0.6.31. |  |