Contrast Coding in Two-Factor Analysis of Variance Studies: An Application to Cotton Data

Abstract

In this study, how the contrast analysis is performed in a two-way factorial design consisting of contrast estimates containing specific questions determined to investigate the specific differences between averages was examined in detail. For this purpose, after determining the hypotheses to determine the main effects, contrast coefficients suitable for each hypothesis were created and contrast analysis was performed. In the study, some of the data obtained from the cotton trials conducted in the Field Crops Department of the KSU Faculty of Agriculture were used with permission. Cotton varieties, years and interaction effects were evaluated using the R and SPSS 21.0 package programs. With the use of contrast, while performing two-factor analysis, first the main effects were investigated and then the interaction effects were investigated. Among the estimates made with 1 degree of freedom within the main effect A, the contrast estimation showed the greatest effect (rcontrast=0.7901). Later, among the estimates made with 1 degree of freedom within the main effect B, the contrast estimation showed the greatest effect (rcontrast=0.6370). Likewise, when looking at the interaction effects, it is seen that the effect of the contrast estimation (rcontrast=0.4388) shown by the quadratic effect of is more important, that is, the quadratic effect is more important. As a result, this study showed the researchers where the main effects were found when the average differences in factorial designs were analyzed and gave detailed information about their effect sizes.