Unsupervised Learning

Modeling Latent Effect Heterogeneity The goal of this project is to provide a series of tools to investigate latent heterogeneity in the effect of treatment variables or other observed covariates. Latent heterogeneity can occur because latent conditioning terms (i.e., interactive factors) are omitted in the empirical analysis. In generalized linear models, omitting interactions can lead to latent occurrences Simpson’s Paradox, which is a long-standing problem in statistical analysis in general and in the social sciences in particular.

Unsupervised Learning Methods and FMM for Election Forensics (with Walter Mebane) Elections are one of the cornerstones of modern democracies. *Election forensics* is a subarea of Political Science that uses statistical methods to investigate frauds in election. Frauds are by their very nature concealed phenonema. The perpetrators don’t want to reveal their act and want to avoid leaving any indication of manipulaiton of the results. This NFS-funded project, with Walter Mebane (PA), proposes some unsupervised learning methods and a series of finite mixture models to estimate the probability of fraud in elections using election data.