Obsidian | Level 7

The details about subject = id(center) in the PROC GEE

Dear all, I met a question in PROC GEE analysis:

I want to add the center as a random effect to the model:

However, in my data, the ID was an independent on Center. (Refers to ID in each enter was not repeated)

`data gee1;input Center ID group Visit1-Visit4;datalines;1 1 1 0 0 0 01 2 1 0 0 0 01 3 1 1 1 1 11 4 0 1 1 1 01 5 1 1 1 1 11 6 0 0 0 0 0 2 7 0 1 0 0 12 8 1 0 1 0 12 10 0 1 1 1 12 11 1 0 1 1 02 12 1 0 0 1 12 13 0 1 1 0 0 ;data gee1;set gee1;Visit=1; Outcome=Visit1; output;Visit=2; Outcome=Visit2; output;Visit=3; Outcome=Visit3; output;Visit=4; Outcome=Visit4; output;run;proc print; run;Proc gee data = work.gee1 desc;class id visit group(desc) outcome(desc) center;model outcome = group / dist = binomial link = logit WALD TYPE3;REPEATED subject = id(center)/ within = visit type = UN CORRW;run;`

And in the example in the https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.4/statug/statug_gee_examples01.htm,

The ID in each center is the same.

Can I use the reapeated subject = id(center) to input center as random effect?

If not, How Can I adjusted for center?

Sincerely hope I can get you help!

1 ACCEPTED SOLUTION

Accepted Solutions
SAS Super FREQ

Re: Deails about repeated subject = id(center) in PROC GEE.

The GEE model does not use random effects, so you cannot add a random effect for CENTER in your PROC GEE model. If observations are considered correlated within IDs and observations from different IDs are considered uncorrelated, then SUBJECT=ID is what you need. And you are already adjusting for CENTER by including it as an effect in the MODEL statement.

2 REPLIES 2
Obsidian | Level 7

Deails about repeated subject = id(center) in PROC GEE.

Dear all, I met a question in PROC GEE analysis:

I want to add the center as a random effect to the model:

However, in my data, the ID was an independent on Center. (Refers to ID in each enter was not repeated)

`data gee1;input Center ID group Visit1-Visit4;datalines;1 1 1 0 0 0 01 2 1 0 0 0 01 3 1 1 1 1 11 4 0 1 1 1 01 5 1 1 1 1 11 6 0 0 0 0 0 2 7 0 1 0 0 12 8 1 0 1 0 12 10 0 1 1 1 12 11 1 0 1 1 02 12 1 0 0 1 12 13 0 1 1 0 0 ;data gee1;set gee1;Visit=1; Outcome=Visit1; output;Visit=2; Outcome=Visit2; output;Visit=3; Outcome=Visit3; output;Visit=4; Outcome=Visit4; output;run;proc print; run;Proc gee data = work.gee1 desc;class id visit group(desc) outcome(desc) center;model outcome = group / dist = binomial link = logit WALD TYPE3;REPEATED subject = id(center)/ within = visit type = UN CORRW;run;`

And in the example in the https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.4/statug/statug_gee_examples01.htm,

The ID in each center is the same.

Can I use the reapeated subject = id(center) to input center as random effect?

If not, How Can I adjusted for center?

Sincerely hope I can get you help!

SAS Super FREQ

Re: Deails about repeated subject = id(center) in PROC GEE.

The GEE model does not use random effects, so you cannot add a random effect for CENTER in your PROC GEE model. If observations are considered correlated within IDs and observations from different IDs are considered uncorrelated, then SUBJECT=ID is what you need. And you are already adjusting for CENTER by including it as an effect in the MODEL statement.

Discussion stats
• 2 replies
• 369 views
• 1 like
• 2 in conversation