True anomaly, published False positive, published Null, filed away

Publication Bias Simulator

This simulator streams a large population of independent scientific studies, each testing whether something anomalous is really there. A small, adjustable fraction genuinely are — the rest are pure noise, and statistical testing guarantees that a fixed share of noise studies will still cross the significance threshold by chance alone. Toggle whether only "significant" results get published, and watch the visible literature's apparent anomaly rate pull away from the true underlying rate as the unpublished file drawer fills up — the same file-drawer and multiple-comparisons dynamic documented across real meta-science and cited to explain inflated positive-result rates in exploratory research.