Oral

An explosion in sophisticated statistical models available to public health researchers has occurred: structural equation models, latent class mixture regression models, generalized multilevel models, multivariate individual change models, and item response theory models, to name a few. All can substantially increase public health researchers’, epidemiologists’ and others’ abilities to investigate increasingly complex questions that more traditional statistical approaches cannot. Moreover, statistical models can allow investigators to address pressing social justice questions. For example, multilevel models allow analysts to simultaneously investigate individual and contextual influences on health, as well as what predicts variance across contexts. Thus, a researcher could chose to examine the role of a neighborhood level factors like local pollutant levels on the health of minority and majority individuals. Likewise, multivariate individual change models could allow one to address whether different trajectories describe individuals’ rates of improvement after intervention efforts, whether different trajectories tend to fall into a smaller set of types, and which variables predict an individual’s trajectory. Thus, an investigator might choose to examine how children’s health changes across time as a function of public health intervention and whether socio-demographic variables predict different trajectories. Unfortunately, the field rarely uses these models because of limited exposure and training. Thus, their power to advance public health and address social justice issues goes unutilized. This session will address that. It will describe different statistical modeling approaches: generalized multilevel models (also known as generalized hierarchical models), latent class mixture regression models, multivariate individual change models, and item response theory-structural equation-based measurement models. Throughout the presentation, panelists will take an applied approach, emphasizing implementation and interpretation over equations. Attendees will leave with a basic understanding of the concepts in each of the approaches, they will know which software packages can estimate the models, and they will have understood an applied example related to social justice. Attendees will have time after each presentation to ask model specific questions, as well as time following the presentation set to ask general questions. Attendees will leave prepared to advance the methodological and quantitative rigor of their studies investigating social justice and other critical public health topics.

Session Objectives: At the session's end participants will be able to: 1. Explain the advantages of a modeling approach over a more traditional statistical approach. 2. Describe the key interpretative features of the models presented during the session. 3. Formulate appropriate research questions for each of the presented statistical models.

CE Credits: Medical (CME), Health Education (CHES), Nursing (CNE), Public Health (CPH)