cancel
Turn on suggestions
Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Browse apps to extend the software in the new JMP Marketplace
Choose Language Hide Translation Bar
GregMcMahon
GregMcMahon
Level III

Predictor screening - question on interpreting the results

Feb 16, 2020 06:16 PM (1444 views)

Hi all,

 

There have been several posts that have been helpful but I am still left wondering about a couple of things. To be clear, this is for the JMP Predictor Screening platform and not JMP Pro Bootstrap Forest.

 

The results of the predictor screening are the result of a bootstrap forest model. The question(s) I have are (and the first has been asked before): What is the contribution value? Is it G squared or SS depending upon data type? Also, the results are based on a bootstrap model, but we have no way of knowing if that model is reasonably good. Can we assume that if all the portions are roughly the same, the model that ran in the background is probably not great (or as was pointed out earlier, there is an important variable missing?

 

Thanks!

0 Kudos
1 REPLY 1
dale_lehman
dale_lehman
Level VII

Re: Predictor screening - question on interpreting the results

Feb 17, 2020 11:23 AM (1401 views)  |  Posted in reply to message from GregMcMahon 02-16-2020

I'd like to add to your question.  Since there is no place for a validation column in the predictor screening platform, I have assumed that it uses all the data.  Is that the main difference between that platform and the bootstrap forest (as well as the lack of parameters to set in the former compared with the latter)?  If so, what is the logic of using a bootstrap forest without validation to screen the predictors, rather than using the bootstrap forest with validation?