Statistical power analysis can be used to increase the efficiency of r
esearch efforts and to clarify research results, Power analysis is mos
t valuable in the design or planning phases of research efforts. Such
prospective (a priori) power analyses can be used to guide research de
sign and to estimate the number of samples necessary to achieve a high
probability of detecting biologically significant effects. Retrospect
ive (a posteriori) power analysis has been advocated as a method to in
crease information about hypothesis tests that were not rejected. Howe
ver, estimating power for tests of null hypotheses that were not rejec
ted with the effect size observed in tile study is incorrect; these po
wer estimates will always be less than or equal to 0.50 when bias adju
sted and have no relation to true power. Therefore, retrospective powe
r estimates based on the observed effect size for hypothesis tests tha
t were not rejected are misleading; retrospective power estimates are
only meaningful when based on effect sizes other than the observed eff
ect size, such as those effect sizes hypothesized to be biologically s
ignificant. Retrospectively power analysis can be used effectively to
estimate the number of samples or effect size that would have been nec
essary for a completed study to have rejected a specific null hypothes
is. Simply presenting confidence intervals can provide additional info
rmation about null hypotheses that were not rejected, including inform
ation about the size of the true effect and whether or not there is ad
equate evidence to ''accept'' a null hypothesis as true, We suggest th
at (1) statistical power analyses be routinely incorporated into resea
rch planning efforts to increase their efficiency; (2) confidence inte
rvals be used in lieu of retrospective power analyses for null hypothe
ses that were not rejected to assess the likely size of the true effec
t, (3) minimum biologically significant effect sizes be used for all p
ower analyses, and (4) if retrospective power estimates are to be repo
rted, then the alpha-level, effect sizes, and sample sizes used in cal
culations must also be reported.