statistical-analysis

To run statistical analysis on SWARM experiment data, execute the 'analyze_csv.py' script and provide the path to your CSV and an output directory as arguments. The script performs data normalization, hypothesis testing, and multiple-comparison corrections.

Statistical tests available for analyzing SWARM data include pairwise Welch's t-tests for hypothesis testing, Cohen's d for effect sizes, and Bonferroni corrections for multiple comparisons. These tests ensure rigorous safety assessment and governance evaluation.

Yes, you need pandas, scipy, and numpy installed to normalize SWARM sweep data. These dependencies are required to load multi-seed experiment data and compute the statistical evaluations correctly.

To calculate effect size and apply multiple-comparison corrections for experimental data, use the Skill to compute Cohen's d and apply Bonferroni corrections. This normalizes data and validates hypothesis testing results accurately.

Yes, this statistical analysis approach works with multi-seed welfare outcome experiments. It loads and normalizes data from SWARM multi-seed sweeps to evaluate different parameters like tax rates using rigorous statistical methods.