Section 9.11 Interpretation
Our logistic regression model predicts the probability that a person will default on their credit card based on balance, income, and student status.
Coefficients:
-
The model found a significant positive relationship between balance and default.
-
As balance increases, the odds of default rise sharply.
-
The student variable had a negative coefficient, meaning students are less likely to default than non-students (about 57% lower odds, holding other factors constant).
-
-
Income showed a very small, non-significant effect once balance was included.
Model Fit:
-
McFadden’s R² ≈ 0.46, suggesting the model explains a moderate amount of variation.
-
The AUC = 0.95, indicating excellent ability to distinguish between defaulters and non-defaulters.
-
Confusion Matrix: high overall accuracy (~97%), but this is inflated due to the low number of defaults (class imbalance).
-
The model is excellent at identifying people who will not default, but less effective at catching rare defaults.
In short:
-
Balance is the strongest predictor.
-
Students are less likely to default.
-
Income doesn’t matter much here.
-
The model performs very well overall.
