1step,rule,n_before,n_after,n_removed,rationale21,drop rows with missing values in required columns,1200,1200,0,complete cases required for OLS32,drop duplicate firm_id,1200,1200,0,one row per firm expected43,drop rows with training_hours < 0,1200,1188,12,negative hours of training per employee are physically impossible; treated as recording errors5