What is a reason we might prefer to use the expected Percent Correctly Predicted

Question

What is a reason we might prefer to use the expected Percent Correctly Predicted (ePCP) to evaluate and compare different logistic or probit regression model specifications?

a. ePCP penalizes model complexity, and thus helps us choose the simplest model possible.

b. ePCP more heavily penalizes false positives (i.e., yˆi > 0.5 when yi = 0)

c. ePCP more heavily penalizes large errors than small errors (e.g., yˆi 0.91 versus yˆi 0.51 when yi = 0).

d. ePCP measures variance in y explained by the model, not classification accuracy.

e. None of these / cannot say

Explanation / Answer

The correct answer is (c) - ePCP more heavily penalizes large errors than small errors (e.g., yˆi 0.91 versus yˆi 0.51 when yi = 0).

Related Questions

Navigate

• Browse (All)
• Browse W
• Subjects

Integrity-first tutoring: explanations and feedback only — we do not complete graded work. Learn more.