PEIRA

Bayes factors use the JZS default: a Cauchy prior on the standardised effect size. The scale sets how large an effect you expect a priori; medium (0.707) is the conventional default.

Blinded interim evidence shows how strongly the two arms differ while an experiment runs — never which way. Hide it if you'd rather see nothing until unblinding.

Everything is stored in this browser only. JSON holds every experiment (schedules stay encoded); CSV exports one experiment's day table for analysis elsewhere.

Each period is drawn by lot. A 1-day period is the classic daily coin flip; longer periods suit slow-acting interventions. Set periods to 0 to run open-ended and stop when the evidence is strong enough.