Relationship of Statistical Tests

All of the statistical tests in the Logistic Regression reports compare the fit of the specified model with subset or superset models, as illustrated in Figure 11.14. If a test shows significance, then the higher order model is justified.

• Whole model tests: if the specified model is significantly better than a reduced model without any effects except the intercepts.

• Lack of Fit tests: if a saturated model is significantly better than the specified model.

• Effect tests: if the specified model is significantly better than a model without a given effect.

Figure 11.14 Relationship of Statistical Tests