cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
JMP is taking Discovery online, April 16 and 18. Register today and join us for interactive sessions featuring popular presentation topics, networking, and discussions with the experts.
Choose Language Hide Translation Bar
anne_sa
Level VI

Fit Model - Custom Test

Hello everybody,

I have a question regarding the Fit Model platform. Let's say that I am working with the Big Class data table. My Y variable is height and I have two effects in my model: age and sex.

I would like to know if the height difference between the groups 12 and 17 for M is significantly different from the height difference between the groups 12 and 17 for F. I think I can use the "Custom Test" feature but I am not sure of how to do that properly...

 

Thanks in advance!

1 ACCEPTED SOLUTION

Accepted Solutions

Re: Fit Model - Custom Test

I just accidentally found this discussion again and realized that I never responded. Yes, you set up the contrasts correctly. You could also use the Custom Test feature to achieve the same result, but it would take more thought to obtain the correct coefficients for the parameters:

 

custom test.JPG

View solution in original post

8 REPLIES 8

Re: Fit Model - Custom Test

Did you search the discussions for this topic first? There is a good discussion that might help you. You can find it here.

anne_sa
Level VI

Re: Fit Model - Custom Test

Hi @Mark_Bailey 

 

Thanks this post is really helping indeed! In my example I have 3 treatments (A,B, C) and 2 Times (T0, T1) and I want to know if the difference between T1 and T0 is significantly different between the treatments.

 

If I understand correctly the first step of your post, in my situation the difference between T1 and T0 (T1-T0 = -T0-T0 = -2 T0) for the treatment A correspond to the following coefficients [0, -2, 0, 0,2,0,0]? I am not sure for the interaction term...

anne_sa_1-1596180073145.png

 

 

Re: Fit Model - Custom Test

You can use contrasts instead of a custom test, although they are equivalent. I think that the contrast is easier to build and interpret. Set them up like this (mock up of data):

 

contrast.JPG

anne_sa
Level VI

Re: Fit Model - Custom Test

Hi @Mark_Bailey ,

 

Thanks for your answer! If I understand correctly, in your example you test that the difference between T0 and T1 is significantly different from 0 for each treatment. So if I want to know if the gap T0-T1 is different between two treatments, I need to subtract the corresponding columns?

But I've probably missed something because I get a warning:

anne_sa_0-1596439113336.png

 

Re: Fit Model - Custom Test

I just accidentally found this discussion again and realized that I never responded. Yes, you set up the contrasts correctly. You could also use the Custom Test feature to achieve the same result, but it would take more thought to obtain the correct coefficients for the parameters:

 

custom test.JPG

anne_sa
Level VI

Re: Fit Model - Custom Test

Perfect, thank you for the confirmation @Mark_Bailey!

And what does the warning mean in that case?

Re: Fit Model - Custom Test

The individual tests are valid but the joint test is invalid. Just don't use the results for the joint test.

anne_sa
Level VI

Re: Fit Model - Custom Test

Ok, thanks again for your help!