Poisson regression with robust error variance - cross sectional study

Question

To estimate RRs in a statistical model with a binary outcome, sometimes the modified Poisson regression can be used (proposed by Zou). Specially in epidemiology, when the incidence rate of the binary outcome variable is above 10%, then it's necessary to use an alternative to the logistic regression because the ORs are no more interpretable as RRs.

My question is about implementing the modified Poisson regression (with robust error variance) for binary a outcome:

Does the use of "repeated subject = id" in the Genmod procedure is appropriate in the context of a cross sectional study?

Answer 1

The modified Poisson regression method suggested by Zou allows estimating the relative risk and confidence intervals by using robust error variances. Using a Poisson model without robust error variances will result in a confidence interval that is too wide. The robust error variances can be estimated by using the "repeated" statement and the subject identifier, even if there is only one observation per subject, as Zou points out.

References:

• http://www.ats.ucla.edu/stat/sas/faq/relative_risk.htm

• Guangyong Zou, A Modified Poisson Regression Approach to Prospective Studies with Binary Data, Am. J. Epidemiol. (2004) 159 (7): 702-706 doi:10.1093/aje/kwh090