hypothesis-testing #

Visualizes the hypothesis testing pipeline: competing statements (H0 vs Ha), a sample producing an observed test statistic z_obs, the null distribution of the statistic under H0, the p-value as the right-tail probability beyond z_obs, and the decision rule comparing p to a pre-specified significance level α (reject H0 when p ≤ α).

canvasclick to interact

⏮◀◀▶▶STEP0.25x1xZOOM

t=0s

practical uses #

01.A/B testing: decide if a new variant improves a conversion rate beyond random noise
02.Quality control: detect if a manufacturing process mean has drifted above a tolerance
03.Scientific experiments: quantify evidence against a null effect using a p-value threshold

technical notes #

Pure Canvas 2D. Uses a 4-stage time cycle (~4.2s) with easing to sequentially reveal sample -> statistic -> p-value shading -> decision. Null distribution is drawn as blocky vertical bars from a standard normal PDF; p-value uses a self-contained normal CDF via an erf approximation. All geometry is grid-snapped for a retro block aesthetic and scaled with scale = min(w,h)/240.

← random-variables continuity →