Turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

- JMP User Community
- :
- Discussions
- :
- Re: Model Comparison with model from PLS has problems

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Dec 11, 2018 7:08 AM
(1357 views)

I've been trying to add a PLS model for binary response to the Model Comparison Platform. After running PLS I save the prediction formulas. This creates 2 columns, one for each level. If you then try to add that model to Model Comparison and specify Pred Formula Y[1] it will ask which column you're trying to predict. Here we enter Y. It will then create a column property, "Predicting" with {"Y"}. At this point when you enter the Model Comparison window you get an error that there is no property indicating what level you are trying to predict.

As a control case I did the same thing with nominal logistic and tried to enter that model into the Model Comparison. Now the column property is "Response Probability" with the formula {:Y, 1, Creator( "Fit Nominal Logistic" ), ID( 546461361 )}. This works fine. I then went back to Pred Formula Y[1] and removed the property "Predicting" and added a new property "Response Probability" with the formula {:Y, 1}. This works fine.

Not sure if I did something wrong or it may just be that whwn PLS models are added you get strange results.

1 ACCEPTED SOLUTION

Accepted Solutions

Highlighted

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Gene:

I took a quick look. Not sure what's going on with the PLS platform. I thought maybe because the Y response was a Data Type: Numeric, Modeling Type: Nominal it might be an issue...but it appears as thought the behavior you cite occurs with changing the Data Type to Character as well. I did use the Multivariate Methods -> Discriminant Analysis -> Wide, Many Columns Discriminant Method to do the analysis. Then publish to the Formula Depot (you can invoke Model Comparison from the Formula Depot report as well...one of my JMP Pro fave platforms), or save columns to the data table, and then invoke the Model Comparison platform and it appears as though JMP Pro behaves as one would expect.

1 REPLY 1

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Gene:

I took a quick look. Not sure what's going on with the PLS platform. I thought maybe because the Y response was a Data Type: Numeric, Modeling Type: Nominal it might be an issue...but it appears as thought the behavior you cite occurs with changing the Data Type to Character as well. I did use the Multivariate Methods -> Discriminant Analysis -> Wide, Many Columns Discriminant Method to do the analysis. Then publish to the Formula Depot (you can invoke Model Comparison from the Formula Depot report as well...one of my JMP Pro fave platforms), or save columns to the data table, and then invoke the Model Comparison platform and it appears as though JMP Pro behaves as one would expect.