Table 2 Assumptions of the estimation methods described in Table 1 that are used for addressing treatment nonadherence in pragmatic trials with point-treatment settings
Name of the method | Key assumption | Other assumptions |
|---|---|---|
ITT, baseline-adjusted ITT, naïve PP, naïve AT | No unmeasured confounding | Consistency, no interference, positivity, nonadherence occurred completely at random |
Baseline-adjusted PP, IP-weighted PP | No unmeasured confounding | Consistency, no interference, positivity, correct model specification |
Naïve 2SLS, first-stage adjusted 2SLS, both-stages adjusted 2SLS | Exclusion restriction | Consistency, no interference, positivity, correct model specification, relevance, monotonicity |
Naïve 2SRI, first-stage adjusted 2SRI, both-stages adjusted 2SRI | Exclusion restriction | Consistency, no interference, positivity, correct model specification, relevance, monotonicity, linearity of residuals |
NPCB | Exclusion restriction | Consistency, no interference, positivity, relevance, monotonicity |