Joint model

[meleseenyew1919]

Hither! can I get the SAS code for The joint modeling of the "longitudinal ordinal and parametric proportional Hazards model" and I'm new to such code so please tell me with an example?

[sbxkoenk · Accepted solution]

Hello @meleseenyew1919 ,

I think you should elaborate a bit more on the data you have at your disposal (or the data you will collect) and on the purpose of your research. Keep it concise but clear.

Also, why are you talking about joint modelling? Are there two targets (target variables) you want to model simultaneously?

I guess you will need PROC GLIMMIX or PROC PHREG (SAS/Stat) but to get a concrete code example, you will need to provide us with more detail. Otherwise the example may be completely wrong.

Cheers,

Koen

[meleseenyew1919]

Hello! I am so sorry I'm too late! to clarify it I have two response variables one longitudinal ordinal and the other survival models. so when I run a separate analysis of each model using the SAS survival model fulfills the assumption of the parametric proportional model and the Weibull distribution is the best distribution and the Longitudinal model also as well. After this, I need a Joint model Analysis for these two so, please every one of you have a Joint SAS code help me.
you can send with out example also I can practice it!

Thank You!

[meleseenyew1919]

Hello, sbxkoenk! sure your guess is correct I need PROC GLIMMIX (SAS) so please help me give me the code even without an example I will practice it!

If you have a question, please!
Thank you!

[sbxkoenk]

Hello @meleseenyew1919 ,

I will post an example in the afternoon (Brussels time). No time right now.

But I think (I am sure) you can find a lot on internet as well on joint modelling of longitudinal and survival time-to-event data with SAS PROC GLIMMIX.

If your data require 'special features' to be built in in the example code, let me know.

Koen

[meleseenyew1919]

Sure I can get a lot of SAS cod but here is the thing the difficulty is that when the longitudinal model is categorical and ordinal to run as jointly!

Thanks for your good communication!

[meleseenyew1919]

Simply the longitudinal submodel is the ordinal response!

[meleseenyew1919]

Hello, sbxkoenk! I am waiting for you.....

[sbxkoenk]

Hello @meleseenyew1919 ,

I know. I should have said eve (Brussels time) instead of afternoon. Sorry for that. Have been busy until now. Will start looking for an appropriate example now. Check back in 30 minutes or so.

Koen

[meleseenyew1919]

ok, thank you very much!

[sbxkoenk]

Hello @meleseenyew1919 ,

As you have probably figured out already you need the 

     dist=byobs(dist)

model option in the MODEL statement of PROC GLIMMIX, where dist is a variable in your data set (having two distinct values for your two targets).

But I don't find an appropriate distribution and link function for your ordinal outcome in PROC GLIMMIX. I guess that's exactly your problem. 

I need to consult a colleague for this.

But are you sure you want to solve this via a 'real' JOINT MODEL?

Because, in principle, the joint modeling of longitudinal and time-to-event data could be considered a form of mixed outcome (outcomes of a different data type). And for mixed outcomes, joint models are only one of the three(?) possible approaches to achieve a 'combined' solution.

I'm sorry you have to wait till tomorrow.

You can also check:

Rizopoulos, D. (2012) Joint Models for Longitudinal and Time-to-Event Data. With Applications in R, Chapman & Hall/CRC,
Boca Raton, FL.

, but it will probably not contain SAS code 🤔😕.

Coming back on this tomorrow.

Good luck,

Koen

[sbxkoenk]

Hello @meleseenyew1919 ,

You marked my first "answer" as a solution. I think that was a bit too "flattered" as your actual question had not been answered at all. When community members see that a question has been solved, they don't look at it anymore. That way you have blocked possible good answers for yourself. But OK, we'll work it out. Unfortunately not today, but hopefully tomorrow.

Have a nice evening,

Koen

[meleseenyew1919]

Hello @sbxkoenk,
I know that you are so busy! And as much as you can you try to help me thanks for that. sure I need it! I think it may be helpful but I can't get the cod there is one research on SAS:
Gilani N. Joint modeling of repeated ordinal measures and time to event data for CHD risk assessment. Biom Biostat Int J. 2019;8(6):204‒212.

DOI: 10.15406/bbij.2019.08.00290
If one can contact with authors or can search with it!

Thank you very much!

[sbxkoenk]

Hello @meleseenyew1919 ,

I will try to find some sample code today.

Is there a PROC GLIMMIX in the article:

Gilani N. Joint modeling of repeated ordinal measures and time to event data for CHD risk assessment. Biom Biostat Int J. 2019;8(6):204‒212.

? (I haven't opened up this article)

If there's some PROC GLIMMIX code in this article, does it meet your expectations or are there shortcomings? If the code there does not meet your expectations (case), explain why.

Thanks,

Koen

[meleseenyew1919]

Hello Koen!
ok if possible in PROC GLIMMIX it fine but most of the
categorical longitudinal outcome uses PROC NLMIXED including this it has
a similar procedure to my analysis. fill free to use PROC GLIMMIX or PROC
NLMIXED I need the output only!

for the article alternative: try it to open
https://medcraveonline.com/BBIJ/BBIJ-08-00290.pdf



Thank you very much!