Log-linear (Poisson) Regression

Log-linear regression is useful for predicting count data (such as the number of defects occurring in a given process) from either continuous or categorical predictors. The summaries and plots produced are analogous to those for standard linear regression.

To perform log-linear regression

Choose Statistics __image\arrow5.gif Regression __image\arrow5.gif Log-linear (Poisson). The dialog for log-linear regression is identical to the dialog for logistic regression, except that the Model - Link field on the Model page has the choices log, sqrt, and identity. For details on the dialog pages, see Logistic Regression.