Föreläsningar

Question 1

Important when testing models

Answer 1

In-sample estimation, and out-of-sample testing

• prevents overfitting in-sample
• checks robustness of model out-of-sample

Question 2

Model adequacy for the logit model

Answer 2

-Standard R^2 does not work; there is an alternative pseudo r^2, for example the McFadden-R^2

• Often used diagnostic for model fit
• -Reciever Operating Charcteristics (ROC) curve and the Arena Under the ROC curve (AUROC)
• -Cumulative Accuracy Profiles (CAP curves) and the Accuracy Ratio (AR)