Predicting the Effect of a Predictor When Controlling for Baseline