Factor Regression Example

Microalpha ships a tiny, fully offline Fama–French three-factor sample under data/factors/ff3_sample.csv. The file contains weekly observations for Mkt_RF, SMB, HML, and the risk-free rate expressed in decimal form. You can run a quick factor attribution against any Microalpha artifact with the utility script reports/factors_ff.py:

python reports/factors_ff.py artifacts/sample_wfv/<RUN_ID> \
  --factors data/factors/ff3_sample.csv \
  --allow-resample \
  --output reports/summaries/factors_sample.md

The script performs an ordinary-least-squares regression of excess portfolio returns on the factor panel, estimating Newey–West HAC standard errors (default lag = 5). When factor and return frequencies differ, returns are explicitly resampled to the factor frequency using compounded returns (no forward-fill). Microalpha’s reporting pipeline automatically incorporates the table into reports/summaries/flagship_mom_wfv.md when the factor CSV is present, so the published walk-forward summary highlights factor loadings and alpha quality.

Because the sample datasets are bundled, no external downloads or API keys are required—ideal for CI environments and reproducible research notes.