Methodological Research Interests
Cost and cost-effectiveness: One of my major methodological interests lies in the exploration of the intersection of causal inference and health policy. Cost data are often subject to right-censoring, which creates some odd complications due to the stochastic nature of cost accrual rates. I seek to address challenges associated with time-varying treatment and confounding in cost and cost-effectiveness analysis, while also dealing with incomplete data and time-updating death risk.
Bounding local average treatment effects: In certain settings, the exclusion restriction assumption invoked by instrumental approaches can be violated. For instance, in a text message-delivered intervention that features both one-way and two-way (interactive) content, the intervention’s effect may depend upon both receipt of the one-way content and extent of engagement with the two-way content. Therefore, the conditional treatment effect (conditional on potential post-treatment engagement) is only weakly identifiable. I seek to understand the bounds on such local average treatment effects as well as the theoretical properties and operating characteristics of relevant sensitivity analysis procedures.
Endogeneity bias: Nonrandom medication use in observational studies not only acts as a contaminant for estimating associations of interest, but renders many modern causal inference methods inapplicable due to the dubious nature of the ‘no unmeasured confounding’ assumption. There are certain modeling approaches based on 1970s structural equation methodology that can deal with these problems quite nicely…if you’re willing to tolerate some parametric assumptions! I have worked on relaxing these parametric assumptions to the extent possible without compromising interpretability.
Applications
I have interests in a wide array of application areas, including (but certainly not limited to) cardiovascular disease, endometrial cancer, neurology, and pediatric infectious disease. My favorite applications often involve disease detection and diagnosis.