-
Cp statistic for a p-parameter regression model is defined asgiven below
- model with lowest Cp value is preferred
- Cp = p is the norm that model is not overfit
- for a model containing the the regressors x1, x2, ....., xk with least square estimators b0, b1, ....., bk, if the regressor variable xk+1 is added producing b0*, b1*, ....., bk*, bk+1*, then
Example