Booker, HelenHouse, Megan2020-03-232020-03-232020-03-10http://hdl.handle.net/10388/12719Traditional approaches to analyzing data generated in plant breeding programs involve the use of ANOVA and an overall average of genotypic performance across environments and/or years. This approach, in some cases, is an oversimplification as it does not handle the unbalanced data that generally accompanies multi-environment and multi-year data well, nor does it allow for heterogeneity of variance. Typically new entries with fewer site years of data are susceptible to large sways in mean values when the check cultivars they are compared to experience unusual values due to the environment. Use of a generalized linear mixed model (GLMM) to predict means using all available information across locations and years results in more accurate comparisons and prediction of superior genotypes. Firstly, the need for a complex modelling system and the benefits of using GLMM will be illustrated using yield of flax cultivars in multi-environment trials. Moreover, using wheat and flax as examples, we will compare the traditional statistical approaches and the mixed model. We will conduct a brief demonstration on the use of R and ASReml software to generate predicted means using a GLMM. Importantly use of the GLMM gives a more representative trait mean for new cultivars, and thus, more accurate comparisons regarding potential performance in commercial fields.enAttribution-NonCommercial-NoDerivs 2.5 CanadaGLMMplant breedingflaxgeneralized linear mixed modelmulti-environment trialsPresentation