- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Dear all,
I'm now handling with longitudinal data analysis.
I am Looking for the code of "joint random effect model" for two ordinal responses.
I also Looking for the code of "joint marginalized model" for two ordinal responses.
does anyone have an example code?
Is there anybody having experience with this statistical methods and SAS program for this analysis?
Accepted Solutions
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Not a bit of experience, so I am just going to throw an idea at you.
Use PROC PRINCOMP to get the first principal component and the "joint score" for each individual. The information about the principal component will tell you how much of the total variability is explained by that component - in other words, how 'jointly' the two measures are related. Once you have this "joint score", you can then proceed to whatever analysis you might choose. This example shows how this was done for a 1 to 10 scale for 14 variables: https://documentation.sas.com/?docsetId=statug&docsetTarget=statug_princomp_examples03.htm&docsetVer... . In your case you have a 1 to 3 scale for 2 variables. To truly separate things, you will likely need a lot of data, but at least that example should get you started.
SteveDenham
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Many results when looking for: sas "joint random effect model" on the web.
What have you tried? Have you looked at proc mixed?
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Yes I did.
But I could not find any practical example for jointing two ordinal responses in longitudinal data.
I do NOT know how to joint these responses!
Do you have the link of example?
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Moved to stats community where you have more chances of getting help.
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
I almost started giving an answer without asking some questions first.
1. When you say a joint ordinal response, my first thought is that you have some sort of panel data (2 endpoints per subject) measured repeatedly. That really doesn't fit the schema for any of the SAS/STAT procedures, but possibly PROC PANEL in the SAS/ETS suite could handle this. That is only a possibility.
2. What are the two ordinal variables? Is there enough levels in either to possibly consider the response as continuous? If that is the case, then PROC GLIMMIX offers some hope for fitting variables with different distributions. However, multinomial distributions are a pain, and I don't know if you can use the approach in the example where a joint model is fit to binary and count data.
3. What do you mean by "joint random effect"? Random effects are not response variables, which is what comes to my mind when I think of "joint" analysis. Are you talking random intercept type models? Clarification of this will really help.
SteveDenham
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Dear Dr SteveDenham;
This is my thesis. I have two ordinal variable which categorized in three levels. These are my responses which shows the severity of disease in my research. I want to build three models and compare them to find the best one. I meant, i need to assessed the joint effect of two measures simultaneously and see the results.
Now the problem is that i can not find appropriate code for jointing my two responses.
I assessed some studied which had the same goal. But there is only the result and codes are not available in articles!
Do you know any similar example for it?
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Not a bit of experience, so I am just going to throw an idea at you.
Use PROC PRINCOMP to get the first principal component and the "joint score" for each individual. The information about the principal component will tell you how much of the total variability is explained by that component - in other words, how 'jointly' the two measures are related. Once you have this "joint score", you can then proceed to whatever analysis you might choose. This example shows how this was done for a 1 to 10 scale for 14 variables: https://documentation.sas.com/?docsetId=statug&docsetTarget=statug_princomp_examples03.htm&docsetVer... . In your case you have a 1 to 3 scale for 2 variables. To truly separate things, you will likely need a lot of data, but at least that example should get you started.
SteveDenham
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content