35.5 Special Considerations

1. Handling Missing Data

• Use generalized boosted models or multiple imputation (Qu and Lipkovich 2009).

2. Violation of Ignorability

Strategies when unobservables bias treatment:

• Use pre-treatment measures of outcome
• Compare multiple control groups
• Conduct sensitivity analysis:
  • Quantify correlation between unobserved confounders and both treatment and outcome to nullify the observed effect

---

References

Qu, Yongming, and Ilya Lipkovich. 2009. “Propensity Score Estimation with Missing Values Using a Multiple Imputation Missingness Pattern (MIMP) Approach.” Statistics in Medicine 28 (9): 1402–14.