An R Shiny application that makes the Bayesian interpretation of statistical significance tangible. Move four sliders — the test's Type I error rate α, its statistical power 1 − β, the prior probability that the hypothesis is true, and a reporting-bias parameter u — and watch the one quantity a reader actually cares about update live: the probability that a published significant result is in fact a false positive (the false discovery rate), shown alongside its complement, the positive predictive value. Each slider is paired with a synced numeric field for entering exact values.
The engine is the Ioannidis–Wacholder framework: among all hypotheses tested in a field, PPV = π(1−β) / [π(1−β) + (1−π)α], with reporting bias reclassifying a fraction of would-be non-significant results into the "significant" pile. A grid of 1,000 hypothetical studies renders the effect directly — hue marks whether a study's conclusion is correct, shade marks whether it was reported as significant, so the red-among-dark cells that a reader of the literature actually sees are exactly the false discovery rate. A 2×2 sensitivity panel makes plain that, at parameter settings typical of clinical trials, the prior and the reporting bias move that rate far more than α does.
The statistical core is a small, pure, unit-testable R layer separated from the interface and organised into labelled modules; every quantity is computed analytically, so no data upload is required. A companion manuscript motivating the tool — and reviewing how the elusive prior probability might itself be estimated — is in preparation.