Планирование экспериментов и рандомизация

★ 7.0 · research

experimental-design is a Claude Code skill that guides the planning of experiments and studies before any data is collected — selecting the right design, performing randomization, applying blocking, and structuring treatment combinations so that effects are estimable and results are interpretable. It covers full and fractional factorial designs, Plackett-Burman screening, central composite and Box-Behnken response-surface designs, Latin hypercubes, crossover, split-plot, repeated-measures, and cluster-randomized schemes, addressing confounding and pseudoreplication at the structural level. Reproducible layouts are generated via Python scripts built on numpy, pandas, and pyDOE3, with Python ≥ 3.10 required. The skill is aimed at researchers and analysts facing questions like "how do I assign subjects to conditions", "how do I avoid confounding", or "what's the best way to test these six factors at once".