I was reading Freidman's book "The elements of statistical learning-2nd edition". Page 365, it talks about partial dependence plot. I don't quite understand how he actually calculates partial depence of f(X) on Xs. Say, I built a model on 4 predictors; I use the model itself as f(Xs,Xc). I want to know the partial dependence of my first predictor V1. If the data looks like :<p> V1 V2 V3 V4 predicted_p <p> 1 1 0 1 0.2<p> 1 2 0 0 0.24<p> 2 1 1 1 0.6<p> 2 2 0 1 0.4<p> Is the partial dependence for V1 has two values, one for V1=1, the other for V1=2? For case V1=1, PD=(0.2+0.24)/2=0.22; for case V1=2, PD=(0.6+0.4)/2=0.5 <p> Do I understand right? Thank you. Besides, are the mean of partial dependence for each predictor equal?