4  Cautions

Things that could make a correct-looking analysis wrong. Read before interpreting results.

• Covariate imbalance after randomization is chance. Do not choose adjustment covariates by looking at Table 1.
• “Cleaned” data is a claim, not a guarantee. The pipeline validates every range and code regardless of what the source says.