Discussion questions#
Does it make sense to calculate the p-value on individual regression coefficients from a ridge or lasso model? Why or why not?
The traditional notion of the bias-variance tradeoff states that when the number of parameters in your model, \(p\), exceeds the number of observations, \(n\), there is no unique solution to your model. How exactly then does ridge regression, which does not remove variables from your model, resolve the bias-variance tradeoff when \(p>n\)?