Subscribe Bookmark RSS Feed

Discriminant Analysis

mashal2k3

Community Trekker

Joined:

Dec 10, 2012

I am a new JMP user, using the stepwise function to select a model, how do I force it use an intercept?

4 REPLIES
pgstats

Community Trekker

Joined:

Aug 30, 2011

Do not confuse stepwise variable selection in discriminant analysis and in regression analysis. In discriminant analysis all variables are considered random. What is estimated is the center position and covariance matrix of each group. An observation is classified into the group which is the closest in terms of the Mahanobis distance, i.e. the most probable group. I don't think the concept of intercept is relevant in that context.

PG

mashal2k3

Community Trekker

Joined:

Dec 10, 2012

Thanks for your response. I guess I left my question incomplete. I use stepwise to select a model, then I would like to apply that model to my data. I should be able to chose to have an intercept or no intercept, at least in the old sas scripts you can do that. Or, do I need another menu to apply the model?

Thanks,

Mashal

pgstats

Community Trekker

Joined:

Aug 30, 2011

After you select your variables, you can select  Apply the Model to get a classification of your data and then choose Score options - save the formulas to create new formula columns next to your data. You may then add lines to your table and the classification will be calculated automatically.

PG

Jeff_Perkinson

Community Manager

Joined:

Jun 23, 2011

To get JMP to fit models without an intercept, you need to turn on the "No Intercept" option in the Fit Model dialog box.

2848_JMP Fit Model No Intercept-3.png

You can do this with the stepwise personality and then, after finding the model, click the "Run Model" button to fit the model you've found.

2849_JMP Stepwise Regression.png

In the resulting report window you'll find a "Save Columns->Prediction Formula" option to save the prediction formula back to the data table to apply the model to the data.

2850_JMP Save Regression Prediction Equation.png

Here you can see the prediction formula has no intercept.

2851_JMP Prediction Formula No Intercept.png

-Jeff