Skip to main content
    Sample Size CalculatorBeta

    What’s new

    What has changed in StatWisely, newest first: new calculators, changes to how the site works, and any correction to a result.

    StatWisely is in beta. Found something wrong? Use “Report a problem” under any result, or write to StatWisely@gmail.com

    • StatWisely is in beta. Every result and every recommendation now has a “Report a problem” link that attaches your inputs, and each result names the engine version that calculated it.
    • The procedure finder shows every answer on the page instead of in menus, and asks only the questions that change the recommendation: how an intervention is assigned (individually, by cluster, crossover, stepped wedge), the study design for a yes/no outcome, paired or independent groups, repeated measurements, the trial aim (superiority, non-inferiority, equivalence) and interim analyses.
    • Reliability and validity studies start from the outcome “Questionnaire or instrument score”, then the score being assessed and the measurement property. Criterion validity against a yes/no diagnosis asks whether the cut-off is prespecified.
    • Paired designs (paired t test, McNemar, paired sign test) can be reached from the finder again.
    • Estimating one mean or one set of category proportions no longer shows the several-groups field unless you ask for it.

    • New measurement calculators: Cronbach’s alpha, the intraclass correlation, Cohen’s kappa and Bland–Altman limits of agreement, each estimated within a margin or tested against a benchmark, and the power of a factor model’s RMSEA test.
    • Time-to-event planning: Cox regression with a numerical predictor, three or more groups, a trend across ordered groups, the one-sample log-rank test, competing risks, a delayed treatment effect and restricted mean survival time; and precision for hazard ratios, survival at a fixed time and median survival.
    • Count outcomes: event rates across three or more groups, a trend, non-inferiority of rates, count regression, and precision for a rate, a rate difference and a rate ratio.
    • The Cochran–Mantel–Haenszel test for a common odds ratio across strata, and survival power over a recruitment schedule.

    • Stratified surveys and finite populations of people or clusters can be planned, and the power a sample size already in hand would reach is available for 27 more tests.