Log-Link and Logistic Regression

Abstract

Regression with log-link or Poisson regression is a model that can be used to study the relative change in an outcome variable. In the log-link regression model, the antilog of each coefficient describes the relative difference in the outcome variable associated with each one-unit difference in the predictor variable. For binary outcomes, the logistic regression model relates the log odds of the outcome with one or more predictor variables. The antilog of each coefficient in a logistic regression model describes the odds ratio of the outcome variable associated with a one-unit difference in the predictor variable.