Support Vector Machine (SVM) KFold no longer available in version 16.2?

[#1]

I have two computers, one with JMP Pro version 16.0 and another with Pro 16.2.  Both in Windows.

In JMP version 16.0, I can use Predictive Modeling>Support Vector Machines>Validation Method>KFold.

With the same data sheet in JMP version 16.2, the KFold option is greyed out.  I have tried many times, with many data sheets, and with many different options, but have never seen the KFold option available for SVM in version 16.2.  Was this capability removed?  How can I perform KFold validation on SVM after updating to 16.2?

[#2]

KFold is still available, but it is never available if the nominal response has more than 2 levels. I am guessing your response variable has 3 or more levels to the response? 

This is because when fitting >2 levels, all pairwise models (so n choose 2 models) are fit and then combined for the overall fit. So the number of pairwise models can get very large, very quickly. If we then did kfold on top of that, the time to estimate could get very long as well as possibly running into memory issues storing all of the results.

That was my initial thought, and then I found that with a continuous Y, KFold is not available, so perhaps your response is continuous. Support Vector Regression was added to SVM in 16, and it seems that allowing KFold for SVR models wasn't enabled. This is likely an oversight (I can't think of a reason why it would be disallowed unlike in the >2 levels classification case), so I'll enter a bug report to get it enabled.

Thank you!

[#3]

Thanks for your quick response.  My Y variable is indeed continuous and I am performing SVR.  My understanding of your response is that KFold for SVR was available (but possibly implemented incorrectly) in 16.0, but not available in 16.2?

Thanks!

[#4]

Yes, it was available in 16.0 (and correctly), but it appears to have been accidentally disallowed by 16.2. I believe it should be allowed, and it should be reimplemented by the next release. I apologize for the loss of functionality and thank you greatly for bringing it to our attention!