My research interest is in propensity score methods. Propensity score analysis (PSA) is a quasi-experimental design used to estimate causality from observational studies. It is generally conducted in two phases:
- Estimate propensity scores (i.e. probability of being in the treatment) using the observed covariates.
- Check balance
- Re-estimate propensity scores
- Estimate effect sizes using typical group differences (e.g. t-tests)
Areas I have worked on:
- Multilevel PSA (see
multilevelPSA
R package)
- Matching with non-binary treatments (see
TriMatch
R package)
- Bootstrapping PSA (see
PSAboot
R package)