Exercise $4$

$($a$)$ Why is $R^2$ not suitable for model selection?

$($b$)$ Instead, we want to use Leave$-$One$-$Out Cross$-$Validation $($LOOCV$)$ to estimate the Mean Squared Error $($MSE$).$ For the given dataset: Calculate hat$(y{)}_1$ by using the training set $(y_2,y_3,y_4)$ and $(x_2,x_3,x_4)$

$\backslash$table$[[x,y$