CHAID Results

Figure 1. CHAID tree with missing independent variable values

For node 3, the value of income level is displayed as >Medium;<missing>. This means that the node contains cases in the high-income category plus any cases with missing values for income level.

Terminal node 10 contains cases with missing values for number of credit cards. If you're interested in identifying good credit risks, this is actually the second best terminal node, which might be problematic if you want to use this model for predicting good credit risks. You probably wouldn't want a model that predicts a good credit rating simply because you don't know anything about how many credit cards a case has, and some of those cases may also be missing income-level information.

Figure 2. Risk and classification tables for CHAID model

The risk and classification tables indicate that the CHAID model correctly classifies about 75% of the cases. This isn't bad, but it's not great. Furthermore, we may have reason to suspect that the correct classification rate for good credit cases may be overly optimistic, since it's partly based on the assumption that lack of information about two independent variables (income level and number of credit cards) is an indication of good credit.