Combining loglinear model with classification and regression tree (CART): an application to birth data

Loglinear model is often used to analyze categorical data. Based on the cell counts of a contingency table, the fitted model presents the association pattern among all variables. This information can be visualized through an association graph. Classification and regression tree (CART) method is based on variation (impurity) reduction for binary splitting; a tree structure is produced as a result of the splitting process. Because these two techniques complement each other, the loglinear model was combined with CART for the analysis of birth data. The analysis was more effective, and the results were more informative since the birth data were analyzed with a combination of the loglinear model and CART.