Hi all,
Thanks in advance for your time. I have a few questions about running a GEE in STATA; I saw a few similar questions posed in now-inactive threads, and hoped to revive the conversation.
I am working on an analysis for which I am estimating the effect of an intervention on a number of behavioral outcomes (in the context of a controlled trial). This data is clustered in two ways: longitudinally (repeated measure, pre-post intervention) and by dyad (e.g., recruitment strategy). I hope to run a GEE to obtain population-level estimates of the exposure on the outcome using STATA, and primarily would like to simultaneously account for both forms of clustering. What I have seen online is that GEE in STATA can only accommodate one form of clustering at a time, but I am told it is possible to account for both. Many others have posted similar questions over the years in this forum without a direct answer, so I am wondering if there is a new workaround I'm not aware of.
My second question regarding a GEE is the feasibility of including a 3-level outcome in the GEE rather than a binary outcome. Is this supported in STATA?
I have seen mention of GLMM modeling instead (using melogit), but I would prefer the GEE model if possible given the population-level estimates and ease of interpretation.
Any advice would be VERY appreciated. Thank you!
Grace
Thanks in advance for your time. I have a few questions about running a GEE in STATA; I saw a few similar questions posed in now-inactive threads, and hoped to revive the conversation.
I am working on an analysis for which I am estimating the effect of an intervention on a number of behavioral outcomes (in the context of a controlled trial). This data is clustered in two ways: longitudinally (repeated measure, pre-post intervention) and by dyad (e.g., recruitment strategy). I hope to run a GEE to obtain population-level estimates of the exposure on the outcome using STATA, and primarily would like to simultaneously account for both forms of clustering. What I have seen online is that GEE in STATA can only accommodate one form of clustering at a time, but I am told it is possible to account for both. Many others have posted similar questions over the years in this forum without a direct answer, so I am wondering if there is a new workaround I'm not aware of.
My second question regarding a GEE is the feasibility of including a 3-level outcome in the GEE rather than a binary outcome. Is this supported in STATA?
I have seen mention of GLMM modeling instead (using melogit), but I would prefer the GEE model if possible given the population-level estimates and ease of interpretation.
Any advice would be VERY appreciated. Thank you!
Grace
