My research focuses on causal inference with observational data, as well as missing data. I have developed a method to assess the therapeutic equivalence of brand vs. generic drugs using observational (i.e., insurance claims) data.
I am also working on deriving the most efficient estimator in settings with non-monotone missing data. These are situations where some of the measurements on patients in a longitudinal study are intermittently missing throughout the follow-up period. For example, patients being treated for additiction are usually required to check in to the recovery clinic at regular intervals. Sometimes the patients will not appear on a check in date, but that doesn't mean they will never check in again. Can we gain efficiency in our estimates by using these patients' data?