Summary information on the specifications used to build the CHAID model and the resulting model
| Model components | Model specification | Results |
|---|---|---|
| Dependent variable | HIV status | HIV+=13.6% |
| Independent variables | Age, age at first sex, marital status, ever had a child, age at first birth, experience premarital childbearing, relationship to the head of household, region of residence, place of residence, education, wealth index, religion | Marital status, age, wealth index, relationship to the head of household, region of residence, place of residence |
| Maximum tree depth | 3 | 3 |
| Minimum cases in parent node | 100 | 100 |
| Minimum cases in child node | 50 | 50 |
| Number of nodes | – | 21 |
| Number of terminal nodes | – | 13 |
| Overall predicted correct percentage | 86.8 |
Twelve independent variables were specified, but only six were included in the final model. The variables such as age at first sex, age at first birth and female education did not make a significant contribution to the model, so they were automatically dropped from the final model. Overall, there are 21 nodes among which 13 terminal nodes. Parent nodes include at least 100 cases whereas child nodes account for 50 cases in minimum
CHAID, χ2 Automatic Interaction Detector.