Table 3.
Comparison of logistic regression models using likelihood ratio test
Model number | Variables added | Model architecture | Residual deviance | df | Model comparisons (LRT) |
---|---|---|---|---|---|
1 | Null | Y ∼ 1 | 2182.1 | 1597 | |
2 | Male | Y ∼ Male | 2149.4 | 1593 | 2 vs. 1: χ2 = 32.70, df = 4, P = 10−6 |
3 | Female | Y ∼ Female | 2168.4 | 1593 | 3 vs. 1: χ2 = 13.63, df = 4, P = 0.01 |
4 | Both | Y ∼ Male + Female | 2136.1 | 1589 | 4 vs. 2: χ2 = 13.29, df = 4, P = 0.01 |
4 vs. 3: χ2 = 32.37, df = 4, P = 10−6 | |||||
5 | Interaction | Y ∼ Male + Female + interaction | 2042.3 | 1573 | 5 vs. 4: χ2 = 93.82, df = 16, P = 10−13 |
Significant LRT signifies a better fit to the data in the more complex model.
LRT, likelihood ratio test.