In this section, we will examine a number of metrics that we can use to evaluate the performance of a linear model other than using the MSE and cross-validation. We will look at some of the statistical tests and metrics that are used to evaluate how well a linear model performs, and to help decide between different linear model forms.
There are two statistical tests to be aware of for linear models, as follows:
- First, is the test for whether one particular coefficient in the model is 0 or not. Failing to reject the null hypothesis indicates that the feature does not seem to contribute much to predictions. The following formulas show these hypotheses:
- Second, is an overall test, that is, the F-test. This tests whether any features have coefficients that are nonzero. Rejecting the null hypothesis suggests that your model has some predictive ability....
