193632 Bayesian Modeling for Various Outcomes in SAS

Tuesday, November 10, 2009

Jichuan Wang, PhD , Epidemiology and Biostatistics, School of Medicine, The George Washington University, Washington, DC
Peichang Shi, MS , Center for Mental Health Services Research, George Warren Brown School of Social Work, Washington University, St. Louis, MO
Nonnormality in data and small sample size are often challenges in public health data analysis. The Bayesian approach does not rely on asymptotic theory, therefore, works very well for various sources of non-normality. More important, with Bayesian approach, small sample inference proceeds in the same manner as if one had a large sample. In addition, using the distributions of the posterior parameter estimates provided by Bayesian analyses, we will be able to estimate the probability that a regression coefficient has a specific value of interest (e.g., the effect of an intervention is positive). SAS 9.2 integrates a BAYES statement in three procedures (GENMOD, LIFEREG, and PHREG) to produce Bayesian modeling and inference capability for various outcome measures, such as continuous, binary, ordinal, nominal, count data, and event history data. This study demonstrates how to use SAS for Bayesian regression model, logit model, negative binomial model, and Cox hazard model using simulated data.

Learning Objectives:
1) Describe the fundamentals of Bayesian approach 2) Discuss the new features in SAS for Bayesian modeling 3) Demonstrate how to conduct Bayesian modeling for various outcomes using SAS

Keywords: Statistics, Behavioral Research

Presenting author's disclosure statement:

Qualified on the content I am responsible for because: I received my PhD from Cornell University in 1990. Since then I have been doing research on public health and related methodology over 18 years.
Any relevant financial relationships? No

I agree to comply with the American Public Health Association Conflict of Interest and Commercial Support Guidelines, and to disclose to the participants any off-label or experimental uses of a commercial product or service discussed in my presentation.

See more of: Statistics Section Poster Session
See more of: Statistics