A general method is described for estimation of the power and sample size o
f studies relating a dichotomous phenotype to multiple interacting loci and
environmental covariates. Either a simple case-control design or more comp
lex stratified sampling may be used. The method can be used to design indiv
idual studies, to evaluate the power of alternative test statistics for com
plex traits, and to examine general questions of study design through expli
cit scenarios. The method is used here to study how the power of associatio
n tests is affected by problems of allelic heterogeneity and to investigate
the potential role for collective testing of sets of related candidate gen
es in the presence of locus heterogeneity. The results indicate that allele
-discovery efforts are crucial and that omnibus tests or collective testing
of alleles can be substantially more powerful than separate testing of ind
ividual allelic variants. Joint testing of multiple candidate loci can also
dramatically improve power, despite model misspecification and inclusion o
f irrelevant loci, but requires an a priori hypothesis defining the set of
loci to investigate.