cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Choose Language Hide Translation Bar
utkcito
Level III

How to save a prediction formula with a different probability threshold?

Hi,

is there a way to do this other than using the Model Classification Explorer Add-In? I am reviewing a few models and it is very time consuming. No matter what I do, either save formulas or predicted or publish, JMP is always using the default threshold it arrived to. I can play with the slider and see the results, but they don't translate to anything outside the report window. Unfortunately that is also the case for the prediction profiler, which I wanted to use with a different threshold.

 

Any insight into this would be greatly appreciated!

 

thanks,

 

uriel

1 ACCEPTED SOLUTION

Accepted Solutions
dale_lehman
Level VII

Re: How to save a prediction formula with a different probability threshold?

Usually the output of a classification model are the probabilities of the classifications.  These are not affected by the probability cutoff - the probability cutoff is applied to these probabilities to produce predicted classifications.  So, the prediction formula you really want is the formula that produces the probabilities.  Then it is a simple matter to produce the classifications from that.  If you want to do this in JMP and save that formula, just create a new formula column with IF (probability>x) then classification A, else classification B (something along those lines) and save that formula.  But the real formula that is of interest is the one that produces the probabilities.

View solution in original post

2 REPLIES 2
pauldeen
Level VI

Re: How to save a prediction formula with a different probability threshold?

An example piece of code that shows what you are doing and what you have so far would help us a lot to solve your issue

dale_lehman
Level VII

Re: How to save a prediction formula with a different probability threshold?

Usually the output of a classification model are the probabilities of the classifications.  These are not affected by the probability cutoff - the probability cutoff is applied to these probabilities to produce predicted classifications.  So, the prediction formula you really want is the formula that produces the probabilities.  Then it is a simple matter to produce the classifications from that.  If you want to do this in JMP and save that formula, just create a new formula column with IF (probability>x) then classification A, else classification B (something along those lines) and save that formula.  But the real formula that is of interest is the one that produces the probabilities.