light-result-analysis

To analyze experimental results for statistical significance, the skill automates per-metric descriptive analytics with 95% confidence intervals, computes effect sizes, and applies omnibus and post-hoc tests with BH-FDR correction to deliver rigorous conclusions.

Generating a publication-ready report from a results.csv file is handled by running the end-to-end analysis pipeline, which coordinates EDA, hypothesis testing, and automated report generation with included templates and visualizations.

Yes, this analysis pipeline supports paired tests and normality checks for multiple metrics by automatically selecting appropriate statistical tests for numeric performance data and computing corresponding confidence intervals.

Leakage screening and data drift detection work during experiment analysis by flagging potential data issues within the reproducible analysis pipeline, ensuring that reported metrics and statistical conclusions remain trustworthy.

Yes, you can use SHAP values and statsmodels for deep-dive experimental result analysis, as the skill leverages these dependencies alongside scipy and pandas to compute effect sizes and generate visualizations.