The main emphasis of this course is on analysing data using the regression methods introduced in STATS 201/208 and their extensions. The practical aspects of fitting linear and generalised linear models are reviewed including estimation, diagnostics and inference. The geometric interpretation of the linear model is presented to enhance the understanding of these models. Simulation-based procedures, including bootstrapping and cross-validation, are introduced as a means to provide robust inference and to investigate the consequences of assumption violations. The two main uses of regression models, prediction and explanation, are discussed. Issues related to predictive models, including ways to estimate prediction error, model selection criteria, model building, and ethics of predictive modelling are discussed. Issues related to explanatory models including confounding, the model choice for causal inference and causal graphs are explored. Other modern regression methods such as lasso and quantile regression are introduced.
Emphasis is on practical application, providing students with a versatile statistical toolbox useful for a range of fields in both academia and industry, including applied statistics, data science, and almost all subjects in Business and Economics, along with any experimental or social science.