plan your sample size.

Pick a test (two-sample / one-sample t-test, one-way ANOVA, two proportions, or Pearson correlation), give any three of effect size, alpha, power, and n, and PowerPlan solves for the fourth with real closed-form scipy power (noncentral t / F, the normal approximation for proportions, the Fisher-z transform for correlation — the same equations as G*Power). A priori power analysis is funder- and IRB-expected, and underpowered designs drive the reproducibility crisis.