Paul D. Allison

Paul D. Allison is a Professor of Sociology at the University of Pennsylvania, affiliated with the Department of Sociology within the School of Arts & Sciences. His research focuses on improving statistical methods for causal inference from non-experimental data, particularly fixed effects models and missing data techniques. He has contributed extensively to Stata modules (e.g., REG2LOGIT, XTDPDML) and authored influential books on logistic regression, survival analysis, and longitudinal data analysis. His work bridges methodological innovation with applications in social sciences.

Education:

• Ph.D., Sociology, University of Wisconsin at Madison, 1976
• M.S., Sociology, University of Wisconsin at Madison, 1973
• A.B., St. Louis University, 1970

Research emphasizes causal inference, panel data analysis, and methodological rigor. Key areas include fixed effects models to control for unobserved heterogeneity, Bayesian approaches for cross-lagged panel models, and maximum likelihood estimation in dynamic panels. He critiques common pitfalls in statistical practices (e.g., lagged dependent variables in mixed models) and advocates for better missing data handling via multiple imputation and maximum likelihood.

His recent work explores sample size requirements for mediation in hierarchical studies, asymmetric fixed-effects models, and the implications of measurement error in panel data. Articles often address practical methodological challenges faced by social scientists.

Notable contributions include textbooks like Fixed Effects Regression Models and Survival Analysis Using SAS, which are foundational for applied researchers. He maintains a prominent statistical methods website and collaborates on interdisciplinary projects, such as analyzing gun violence risk and health outcomes.