Research
Exact, Nonparametric Sensitivity Analysis for Observational Studies of Contingency Tables
This paper develops the first exact, nonparametric sensitivity analysis framework for multi-category contingency tables, providing valid inference under unmeasured confounding. It generalizes Rosenbaum’s sensitivity analysis to general I×J treatment–outcome structures, with both exact and large-sample methods.
Towards Robust Matched Observational Studies with General Treatment Types: Consistency, Efficiency, and Adaptivity
This project extends sensitivity analysis to matched observational studies with general treatment types, investigating optimal test statistics through design sensitivity and Bahadur relative efficiency.
Sensitivity Analysis for Matched-Pair Cluster Randomization Designs
This work develops a class of cluster-level rank tests — mean, sum, and size-weighted — for matched-pair cluster randomization designs, and derives their design sensitivity under lighter assumptions than existing work: unequal cluster sizes and independence rather than i.i.d. clusters.
Sample Size Calculation in Causal Inference for Unmeasured Confounding
This work derives the closed-form number of matched pairs required to remain insensitive to unmeasured confounding up to a prespecified bias level, establishes a central limit theorem for the sensitivity value Γ*, and introduces a finite-sample correction so a study can be designed to control both Type I and Type II error under hidden bias.