Epidemiology, health policy and outcomes
Kaleb Michaud, PhD
University of Nebraska Medical Center
.jpg "Todd Schwartz, PhD photo")
Todd Schwartz, PhD
University of North Carolina-Chapel Hill
This session is intended to provide an overview of commonly used statistical modeling strategies to analyze independent observations.Rheumatology studies can involve a variety of types of response, or dependent, variables, based on the focus of the study. These variables can be generally classified as continuous (normally distributed), continuous (counts), binary, ordinal, polytomous or time-to-event. Each has a commonly used modeling strategy: ordinary least squares linear regression, Poisson regression, logistic regression, cumulative logit regression, generalized logit and Cox proportional hazards regression, respectively. These various modeling approaches share many similarities, but they also have important distinctions which impact their interpretation and applicability to research questions. This session will focus on choosing an appropriate modeling strategy. Faculty will describe the salient features of each model. They will tailor the session to be relevant for participants through extensive use of examples from the field of rheumatology, placing particular emphasis on the interpretation of these results in peer-reviewed publications. This session will demonstrate the direct connection between the models and the interpretations that may matter most to clinical researchers.
