Choose Language Hide Translation Bar
Community Trekker

Stepwise for Multiple Responses



Let's say I have:

Factors : A, B, C

Responses : Y1 , Y2


If I use Fit Model Personalities : Standard Least Square.

Then the model for example provide : Effect Summary that shows the effects of terms of factors vs both response Y1 and Y2 (in one summary).


However, if I use Fit Model Personalities : Stepwise.

Then, the JMP provide me individual Effect Summary for each response Y1 and Y2.

Meaning, I get two effect summary here. One for Y1 , and one for Y2.


In the case of Standard Least Square, I get only one Effect Summary (which I assume is for combination of Y1 and Y2 responses).


My questions:


1. Both method in the end will provide a Profiler, but they are both different (for sure). Which is more accurate or should we follow?


2. Is there a way for Stepwise to provide one single Effect Summary that consist the whole effect terms vs all responses?






0 Kudos

Re: Stepwise for Multiple Responses

The feature of Fit Model to specify more than one response is convenient but I recommend fitting each response separately and saving the selected model. Stepwise is a great way to accomplish this task. After using Stepwise for each response, click Run Model in the control panel for a given response. Now you are back in Fit Least Squares. Click the red triangle at the top and select Save Columns > Prediction Formula. Repeat this process for each of the processes. When all of them have been been fit and saved, select Graph > Profiler. Select all the Pred Formula columns for this platform. Now you can work with them together.


You don't need to use a fitting platform (and fit the model over and over again) in order to use the Prediction Profiler.

Learn it once, use it forever!
Community Trekker

Re: Stepwise for Multiple Responses

Thanks Mark.


I found out that you could also click run for each of the response in the Stepwise window, and it will add into the same model window.


Then click on the red triangle to get the profiler.


I checked the results and the profiler between the above method and yours are exactly the same.

0 Kudos