Bio

'All models are wrong, but some are more useful than others. Since all models are wrong the scientist must be alert to what is importantly wrong. It is inappropriate to be concerned about mice when there are tigers abroad.'' George Box I have a passion for analyzing social phenomena using methods at the intersection of data science, econometrics, and qualitative research. Data science can play a role in understanding and solving difficult social problems - but it requires deep understanding of the social and institutional context in which the data are gathered. A single record in a database can represent a tremendously complex social interaction. People can voluntarily interact with a system or service (a record in a general physician's healthcare system) or in some cases it can be forced upon them (an arrest record in a police database). They may do so in order to address some underlying problem or need. The service or treatment they receive may or may not be effective. Whether or not the data actually gets entered may itself depend on many institutional or behavioral factors. Parameters in a model may be picking up on any one of these potentially contradictory signals. Without understanding the individual and institutional actors and social forces that generate data, data analysis is liable to produce naive or misleading results. Skills: Data science: regression modeling approaches (OLS, GLM, Cox); LASSO, penalization, variable selection; trees, random forests, boosting, and deep learning; unsupervised learning (clustering with K-Means and K-Prototypes); data ethics and ML equity. Causal inference: randomized experiment design; propensity score matching; instrumental variables; regression discontinuity; differences-in-differences; synthetic control; entropy balancing. Data engineering: ETL and data pipelines; business intelligence and reporting; data architecture.