Mallow's Cp is a statistic used to help with model selection within linear regression. It tests all possible subsets of predictors from the true model to find which combinations are optimal. The formula is given as:
$C_k = \frac{SSE_k}{S^2} - n +2k$
I want to know what the general trend is in $C_k$ when a regressor variable is added or removed from our model. Is there a general trend? Will $C_k$ become smaller generally when we remove variables? How could we use the formula to demonstrate this idea?
Many thanks.
