Parametric and semiparametric methods are widely used in epidemiology. Common examples of parametric models include logistic and normal linear models. Common examples of semiparametric models include the Cox proportional hazards model and the restricted moment model commonly estimated via generalized estimating equations. We will discuss and contrast advantages of each framework. To clarify the role of semiparametric thinking in epidemiology we will work through examples illustrating parametric and semiparametric estimation using standard software (e.g., SAS). Examples will be based on publically-available data from a randomized trial (Hammer, et al NEJM 1997), and computer code made available to attendees.
The workshop will be organized with a 90-minute session covering parametric models (Cole et al AJE 2014), a break, then a 90-minute session covering semi parametric models, and will conclude with a discussion.
Workshop Instructor: Stephen Cole and James Robins