Abstract
In epidemiological studies, sample size requirements are determined on the basis of the desired level of ststistical power. The search for interaction effects is common but the power of such studies is a major concern. This is a particular concern as it is likely that many future studies will be planned to examine gene-gene and gene-environment interactions. In this paper, we consider the situation of a continuously distributed outcome with a categorical or continuous environmental exposure and a categorical genetic factor. We formulae the sample size required with relevant parameters. Also, we consider how the sample is affected if there is misclassification in the assessment of the genotype and the measurement error in the environmental exposure in the model.