rqlm: R package for implementing the modified Poisson and least-squares regression analyses

Abstract

Logistic regression has been widely used for multivariate analyses of binary outcomes in clinical and epidemiological studies. However, the odds ratio is not a directly interpretable effect measure, and is only interpreted as an approximation of risk ratio when the event frequency is low. As effective alternative methods, the modified Poisson regression (Zou, 2004; Am J Epidemiol 159, 702-706) and the modified least-squares regression (Cheung, 2007; Am J Epidemiol 166, 1337-1344) have been standard multivariate analysis methods in recent clinical and epidemiological studies. These methods provide risk ratio and risk difference estimates, which are directly interpretable effect measures regardless of the frequency of the events. The rqlm package involves computational tools for the analyses using the modified Poisson and least-squares regressions with simple commands. This article provides a tutorial for the rqlm package involving example R codes.