Bootstrap without replacement
cratto opened this issue · 1 comments
cratto commented
The help for prtDataSetClass.bootstrap says it uses sampling with replacement. I would like to randomly subsample my training data without getting redundant observations. Does that functionality exist in the PRT?
kennethmorton commented
Surprisingly it didn't. So I added it.
In the devel branch there is new method of prtDataSetBase, bootstrapUnique. This should do what you need. It has been tested but hasn't been run through the ringer a lot. Let me know if you have any issues.