Covariate Balance Assessments for Matched Datasets
In my randChecks R package I provide randomization tests and graphical diagnostics for assessing covariate balance in matched datasets. In short, randChecks uses randomization tests to test whether or not subjects in a dataset are effectively randomized according to a particular experimental design (e.g., complete randomization, block randomization, or rerandomization). See this tutorial for randChecks, which is available on CRAN. Methodological details are in Branson (2021).
Gaussian Process Regression for Regression Discontinuity Designs
In Branson et al. (2019), we propose to use Gaussian process regression for regression discontinuity designs. In short, we fit two Gaussian process regressions (one on either side of the discontinuity), extrapolate both to the discontinuity, and estimate the treatment effect as the difference of these two extrapolations. We take a fully Bayesian approach - we place priors on the covariance parameters within the Gaussian process regressions, thereby propagating uncertainty in the smoothness of the treatment and control response surfaces. In Rischard et al. (2020) we take an analogous approach for geographic regression discontinuity designs.
If you would like to implement our approach, see these replication files on my Github page. Everything is implemented in R and Stan. If you have any questions, please feel free to reach out (and thank you for your interest in our approach!)