Statistical power calculator

This is planning arithmetic, not a replacement for a statistics review. The t tests run the noncentral t distribution under Student's equal-variance assumption; two proportions run Cohen's arcsine method, a normal approximation; correlation runs the Fisher z normal approximation; ANOVA runs the noncentral F with equal cell sizes and equal variances assumed. None of that covers clustered, longitudinal, or covariate-adjusted designs, unequal variances, or multiple comparisons, and the solver cannot warn you that an effect size is unrealistic. Achieved power lands slightly above the target because sample sizes are whole numbers. Computed in this tab; nothing is uploaded.