Idea: because the OLS estimates are chosen to minimize the sum of squared residuals, the SSR always increases when variables are dropped from the model.

• 由于OLS是用于最小化残差平方和；当有变量被从模型中舍弃时；SSR必定上升.