Publication Bias & the Funnel Plot (2D)

The scientific method promises that experimentation and analysis converge on the truth — but only if every result that gets run also gets seen. This simulator draws a batch of independent studies that all measure the same real effect, each with its own sample size and sampling noise, computes each study's significance the way a real researcher would (z-score and two-tailed p-value against the null hypothesis), and then filters which ones get "published" using an adjustable file-drawer strength and significance threshold. Three linked panels show the mechanism from every angle: the funnel plot (effect size against precision, pannable and zoomable) goes lopsided as bias increases; the Egger's regression panel exposes the actual asymmetry test and derives a bias-corrected effect estimate from the same fit; and the running-estimate panel shows the naive pooled effect drifting away from the true value as more biased studies accumulate — exactly the mechanism behind real-world reproducibility problems in science.