|
Abstract:
|
In analysis of variance (ANOVA) for genetic markers, the classical ANOVA methods may not be directly applicable. In humans, though we can treat the paternal and maternal alleles as two separate risk factors, the marker genotypes are often incomplete because their phase information (i.e., the origin of the parental alleles) is missing from most of the current genotype typing technologies. In this study, using a revised Fisher's ANOVA model, we examine estimation and hypothesis testing of the variance components for genetic markers with unphased genotypes based on a random sample from a study population. We show that this revised ANOVA model can retain the same partition on the genetic variance components as the traditional Fisher's ANOVA model. We clarify the difference between the hypothesis tests of the fixed effects in a classical dummy-variable based general linear model (GLM) and the variance components in the revised ANOVA model. We point out that this revised ANOVA model can reduce the confounding between the additive allelic effects and allelic interactions. As a result, it could be more beneficial than the classical GLM for detecting allelic interactions.
|
