Chen-Pin Wang, Ph.D.
I have 12 years of experience conducting research using electronic medical records (EMR). I am the PI on two NIH funded studies that assessed health outcomes in nationwide veterans with type 2 diabetes. I also contributed to analyses of numerous studies involving EMR from VA and Department of Defense. I have developed the expertise in data validation, handing missing data, modeling of comorbidities and heterogeneity and causal modeling for longitudinal observational studies. I have (1) derived the asymptotic relationship between Bayesian posterior predictive p-value and information criteria for model fit; (2) derived a Kullback-Leibler based information criterion for comparing both nested and non-nested models; (3) developed a comprehensive residual diagnostic procedure for assessing model fit of general latent variable (mixed effects) models; (4) extended the utilities of propensity scores to health disparity research. Statistical Modeling of Longitudinal Studies - Latent Variable Modeling (LVM) is an advanced statistical modeling technique for assessing repeatedly measured outcomes. LVM provides a robust parametric modeling framework by employing both discrete and continuous random variables to characterize complex correlation, non-linear correlation, or non-normal outcomes. I have developed expertise in the advancement of LVM, including (1) employing the Bayesian technique to overcome non-identifiability in estimation of competing risks models; (2) integrating the pseudoclass imputation technique with empirical Bayes estimation to obtain patient-level estimates; (3) integrating propensity score technique with the principal stratification methodology to derive causal inference and mediation effects for comparative effectiveness or pharmaco-epidemiology studies.
DEB Assistant Professor
Ph.D., Statistics, The University of Florida, 1999
M.S., Statistics, The University of Florida, 1995
B.S., Mathematics, Central University, 1993
Phone: (210) 617-5300