DPH 707 equips doctoral students with the statistical toolkit needed to analyse public-health data whose outcomes are not normally distributed, such as yes/no responses, counts of clinic visits or rates of infection. The course starts by strengthening core probability concepts (random variables, expected value, variance and the laws of large numbers and central limit) that underpin modern inference. Building on that foundation, students explore the family of generalized linear models, learning how logistic regression handles binary outcomes, how Poisson and negative-binomial models describe counts and rates, and how multinomial or ordinal models address multi-category responses. Through hands-on exercises in R/SPSS (with brief comparisons to SAS and Stata), they fit these models, diagnose assumptions with residual plots and influence statistics, and deal with common hurdles such as over-dispersion or clustered observations using robust and quasi-likelihood methods. Case studies drawn from vaccine-effectiveness surveys, injury-incidence registries and policy-impact datasets guide learners from raw data to clear, policy-relevant conclusions. By semester’s end, participants can choose the right model for a given outcome, interpret effect sizes and uncertainty, and communicate results, and their limitations, in language accessible to clinicians, policymakers and multidisciplinary research teams
None