Rendered Final Report

Rendered HTML report of the full analysis, kept in sync with main. Latest addition: directly tests whether controlling for COVID severity resolves the beta4 placebo failure (Section 4.3) -- it largely doesn't (8% shrinkage for food service, retail's effect actually grows). COVID severity is a real, correlated factor behind the exposure-gradient confound, but not the full explanation; the report is revised to state this honestly rather than overclaim the earlier correlation finding.