Overview

This question evaluates understanding of logistic regression assumptions and limitations for Probability of Default modeling — covering linearity in the log-odds, predictor interactions, multicollinearity, class imbalance and rare events, probability calibration, and the interpretability versus flexibility trade-off within the Machine Learning and credit risk modeling domain. It is commonly asked to assess a candidate's ability to recognize theoretical constraints and practical implications when choosing and validating risk models, testing both conceptual statistical understanding and practical application in real-world, imbalanced-data and regulatory contexts.