Wald test vs linear combinations testing

I am trying to decide which independent variables to remove from a multiple regressoin model in order to improve the RMSE. Would it be better to test for linear dependence between independent variables or use a Wald test to see which variables to drop.