experiment-design

Reproducible experiments require fixed random seeds, environment recording, and versioning. This skill generates those setup checks alongside data splits and cross-validation templates to ensure your Python analysis can be fully replicated later.

Pre-registration records your analysis plan before running experiments to distinguish confirmatory from exploratory findings. This skill provides pre-registration templates that clearly label results, preventing post-hoc bias in your reporting.

Power analysis estimates the sample size needed to detect an effect reliably. This skill generates power estimates using statsmodels and numpy, integrating them into your experimental design to establish statistically valid stopping rules.

Yes, this skill standardizes experimental controls by generating train/test splits and cross-validation configurations for scikit-learn. It structures these controls within a reproducible framework to maintain consistency across data science experiments.

Yes, this skill requires a Python environment with numpy, pandas, scikit-learn, and statsmodels installed. These dependencies power the data splitting, power analysis, and environment recording needed to generate reproducible experiment templates.

Reproducibility fails without environment recording and versioning alongside fixed seeds. This skill addresses this by standardizing the entire experimental design, adding data splitting rules, preregistration, and reporting templates to ensure full replicability.