Ayres Abstracts
| Ayres, M.P., and D.L. Thomas. 1990. Alternative formulations of the mixed model ANOVA applied to quantitative genetics. Evolution 44:221-226. pdf |
| Many recent studies in the field of quantitative genetics draw on results from mixed-model factorial experiments (Ayres et al., 1987; Futoyma and Philippi, 1987; Grosberg, 1987; Hare and Kennedy, 1986; James et al., 1988; Mazer, 1987; Pashley, 1988a; Petranka and Sih, 1987; Shaw, 1986; Stephenson and Winsor, 1986; Via, 1984). But the analysis of such experimental data is confounded by discrepancies in the recommendations of various authors. Some widely used computer packages (e.g., SAS [SAS Institute, 1985] and BMDP [Dixon, 1985) employ different algorithms, which are derived from different statistical models and are based on different sets of assumptions (Hocking, 1973). These alternative models lead to different expected mean squares. Because it is the expected mean squares that determine the appropriate denominator for conducting F tests, different expected mean squares lead to different F tests and can alter important biological conclusions; estimates of genetic variance and heritability are especially sensitive to the choice of model. Here, we identify features that distinguish the models, evaluate the effect of model choice on two recent studies, and offer some recommendations for future research. |
| Mixed model ANOVA/ Heritability/ Reaction norms/ Expected mean squares/ Additive genetic variance |