Monotonicity Assumptions in Deriving Bounds for Generalization

• In Event: Poster Session 15
  In Poster Session: Accuracy and Meaning in Modeling

Abstract

Statisticians have recently developed propensity score methods to improve generalizations from randomized experiments that do not employ probability sampling. However, these methods require strong and often controversial assumptions, which affect the credibility of inferences. This article considers an alternative assumption, monotone sample selection, that partially identifies the population parameter, yielding a range of plausible values in place of a point estimate. While this assumption cannot be empirically validated, we discuss its plausibility in practice. We derive the bounds under this assumption and explore how the bounds are tightened using post-stratification on estimated propensity scores with observable covariates. Results from a simulation study examine the types of covariates that yield the largest precision gain.

Author

• Wendy Chan, The University of Pennsylvania