In my original dataset, the response variable (RESIDUALS) can be less than zero. However, all the predictions (P_RESIDUALS in data_out) are non-negative. Where is the setting to allow negative predictions?
proc hpsplit data=data_in nodes;
class DIM1 DIM2;
model RESIDUALS = DIM1 DIM2;
grow RSS;
prune REDUCEDERROR(leaves=8);
output out=data_out;
id _Leaf_ DIM1 DIM2;
run;quit;
Try removing the Leaves= on the PRUNE setting or even test without pruning.
Just how common are your negative values? If they are uncommon I would suspect that you may need a fuller tree to get them.
Try removing the Leaves= on the PRUNE setting or even test without pruning.
Just how common are your negative values? If they are uncommon I would suspect that you may need a fuller tree to get them.
It's finally time to hack! Remember to visit the SAS Hacker's Hub regularly for news and updates.
Check out this tutorial series to learn how to build your own steps in SAS Studio.
Find more tutorials on the SAS Users YouTube channel.
Ready to level-up your skills? Choose your own adventure.