Classifier analysis in JMP11.

Question

Using JMP11, I want to find a classifier from about 100 variables, which would explain which if the 30 samples falls into one category or into the other (0 or 1). So far I am using stepwise modeling from the Fit model menue. I would like to see which sample gets classified correctly, and which not. Where do I see this in the report? Also I would like to be able to apply the algorithm to other datasets. Where is this option? Thank you for yo time.

michael_jmp · Answer

Sure. To do that, you would save the Probability Formula in the Logistic regression report red triangle menu. That option will save multiple columns to your data table, but you will be interested in the “Most Likely Y” column (where Y is the name of your response variable).

After you have the most likely column, you can find the ones that are misclassified using row selection. Rows menu -> Row Selection -> Select Where. In the Select Rows window, check the Compare column checkbox and choose the original response column on the left and the most likely column on the right. Make sure to change the middle dropdown to “does not equal” before you click OK.

Best,

Michael

andreas_scherer · Answer

Thanks, Michael. From that confusion matrix, there must be a way of extracting information on which sample was misclassified, I am sure.

michael_jmp · Answer

Hello andreas.scherer,

If you have run your model in Stepwise, you should be able to click the "Run Model" button in Stepwise to get a Logistic Regression report for the model chosen in Stepwise. From the Logistic red triangle menu, you can select Confusion Matrix to get a matrix of observations that were correctly and incorrectly classified.

Best,
Michael