Statistics for Risk Modeling (SRM) Conceptual Practice Exam

Mallow's Cp is a statistical tool used in model selection, particularly to assess the trade-off between the complexity of a model and its fit to the data. The primary aim of Mallow's Cp is to provide an estimate of the test mean squared error (MSE) while accounting for the number of predictors in the model.

The statement that Mallow's Cp is an unbiased estimate of the test MSE when calculated using an unbiased estimate of σ² is accurate. An unbiased estimate of σ² is derived from the residuals of the model and reflects the true variance of the errors. When Mallow's Cp employs this unbiased estimate, it provides a more reliable measure for evaluating the model's predictive performance.

In contrast, if a biased estimate of σ² is used, Mallow's Cp could lead to incorrect inferences about the model's performance and potentially mislead model selection. It is also essential to recognize that while Mallow's Cp is helpful, it is not solely indicative of low test error, especially when used alone without consideration of the underlying context or other model evaluation metrics. A small value of Mallow's Cp does suggest a relatively good model fit compared to the number of predictors used, but by itself, it does not provide a