CHOOSING AN APPROPRIATE CUTOFF POINT BASED ON THE FALSE-NEGATIVE AND FALSE-POSITIVE INTERSECT

The motivation for the present work is to search for a more appropriate cutoff point for the purpose of improving predictive accuracy in logistic regression models. In comparison with many previous studies, where the default rate may possibly be underestimated, our findings are more robust. Likelihood ratio testing, multicollinearity diagnostics and goodness-of-fit testing are utilized to search for a better logistic regression model. Our proposed approach, using the cutoff rate where the false-negative rate curve intersects the false-positive rate curve, gives better performance than using 0.5 or the total predictive error rate as the cut-off point. An empirical example is utilized to demonstrate the effectiveness, efficiency and robustness of the proposed techniques.