In this note, we show how to construct the well-known preliminary test estimator to be practical in the high dimensional linear regression model. The strategy is to use a high dimensional test when it is a priori suspected that the parameter may be equal to zero.