Subscribe Bookmark RSS Feed

Repeated Measures Longitudinal Regression in JMP Pro 13

masandberg

Community Trekker

Joined:

Sep 2, 2014

JMP Pro 13 allows assigning up to two Subject terms for the Exchangeable Structure of the Repeated Structure tab in the Mixed Model dialog.  Can anyone give me an example where two subject terms would be appropriate or preferred for repeated measures longitudinal regression?  In particular, I'm wondering whether I could use ID and Eye as Subject terms for the partial dataset listed below (where Eye is nested within ID) and Year_Cat as the repeated term.  Or should I use only ID as the Subject term and concatenate "Eye" and "Year_Cat" as the Repeated term?  Thanks in advance.

 

IDEYE (right=1, left=2)YEAR_CAT
58910
58911
58912
58913
58914
58920
58921
58922
58923
58924
67010
67011
67012
67013
67014
67015
67016
67020
67021
67022
67023
67024
67025
67026
5 REPLIES
jiancao

Staff

Joined:

Jul 7, 2014

As an example, in measuring the effects on achievement scores of students where students are nested within schools, two subject terms--school and students within a school--would be appropriate. 

In your case you can specify the random subject effects as shown below and leave the default Residual for Repeated Structure:

 1-5-2017 4-17-24 PM.png

masandberg

Community Trekker

Joined:

Sep 2, 2014

Yes I understand that. But I have been advised to use an exchangeable structure and avoid using random effects code. So can I use both ID and Eye as Subject terms and Year_Cat as the Repeated term, all within the repeated structure tab? Or should I use only ID as the Subject term and concatenate "Eye" and "Year_Cat" as the Repeated term? The latter is what Julian Parris had recommended to me for JMP Pro 11 a couple of years ago.
jiancao

Staff

Joined:

Jul 7, 2014

 

 

 

masandberg

Community Trekker

Joined:

Sep 2, 2014

"The difference between the two options is how measurements taken from right eye and left eye should be characterized. I think they are from different subjects rather time points."

Each ID has data from two eyes (or only one eye if data from the other eye are missing). And each eye of a given ID has multiple time points. For example, ID 1 might have right eye — year 0, year 1, year 3, and year 5 — and left eye — year 0, year 1, year 3, and year 5.

So which of your two options would be preferable for me?

Thanks,
Michael
jiancao

Staff

Joined:

Jul 7, 2014

If I believe the difference in the measurement between the left eye and right eye is random then I would use the first option. If not, then the second.