I. True/False/Uncertain - Briefly explain.

. 1. Cross validation is used to avoid overfitting.

2. The model averaging procedure would use all the data to estimate the coefficients.

3. Logit model is preferred to Probit model.

4. The higher the cutoff value for classification, the higher the specificity.

5. Cumulative Logit model is preferred to Multinomial Logit.

6. The Negative Binomial regression has more parameters to estimate than the Poisson.

7. Cluster analysis seeks to group similar variables together.

8. Replacing missing values with sample means reduces data variability.