For the time variable, the main effect tests the hypothesis that outcome measure differs between short and long term assessments while controlling for baseline scores and other covariates, including trial arm (that is, testing the effect of time on the outcome measure, while the effect of trial arm and all other covariates held constant).
In Table 5 we present P values, unadjusted and adjusted for multiple comparisons, for the main effects tested, beginning with single-locus associations.
The figures depict the main effects tested by the linear models, with each variable (continuous vs rest-inserted loading, 100 vs 1000 repetitions) depicted in a separate figure.
P-values were obtained using the "Anova" function of the "car" package [ 32] with type II sums of squares (main effects tested without removing variation attributed to interaction terms).
After checking overall data consistency (graph plot) and distribution type (Kolmogorov-Smirnov normality test) of ICT with the UNIVARIATE procedure, values were submitted to analysis of variance with the PROC GLM statement for main effects testing.
Given that many previous candidate gene studies, and even many GWASs, were powered to detect only main effects, testing these samples for gene-gene and gene-environment interactions has the potential for false positive and false negative results [ 2, 3].
The power of the main effect test for treatment was examined.
Significant interactions with treatment were analyzed with simple main effect test followed by Fisher LSD where indicated.
Empirical results show that the presence of epistasis can increase power to detect the underlying loci compared to the non-epistatic additive case, in two-marker full model tests as well as in single-marker main effect tests.
In case of significant interactions, simple main effect tests were performed; that is, the effects of one factor holding the level of the other factors fixed.
