A practical guide to increasing statistical power in research studies with limited participants. The core insight: use a within-subject design (each participant does multiple conditions) combined with multiple tasks per condition, treating outcomes as semi-independent data points with a crossed statistical correction (Clark's min F'). A between-subjects design gains almost nothing from adding more tasks, but within-subjects designs show meaningful power gains. The author introduces PIPS (Project Impossible Power Simulation), an open-source Monte Carlo simulator runnable in-browser via GitHub Pages, built to help researchers estimate how many participants and tasks they need. Key caveats include ensuring task independence, managing learning/contamination effects in within-subject designs, and the tool's lack of peer review.
