The availability of different designs and tests to detect QTLs in associati
on studies raises questions about the relative merits of the various approa
ches. We therefore compared the power of quantitative versus categorical te
sts, the power in population samples versus samples with subjects selected
on the basis of their trait scores, and the power of tests that control for
population stratification using parental genotypes versus tests that do no
t control for stratification. In case-control samples the power of quantita
tive tests was clearly better than that of categorical tests especially whe
n the control group was a population sample. In samples of genotyped trios
of cases and their parents, the power of quantitative tests was much poorer
. Compared to population samples, selection always improved the power in ca
se-control samples where the controls were sampled from the opposite end of
the continuum and frequently deteriorated the power when the controls were
a population sample, Mainly because subjects with at least one heterozygou
s parent need to be selected, the use of tests that control for stratificat
ion resulted in a substantial decrease of power. In the final section our p
ower calculations were integrated into a more general discussion about opti
mizing designs in association studies.