GEE vs mixed models for cross-sectional sibling data
GEE vs mixed models for cross-sectional sibling data
Loading saved threads...
Aaliya Ahamed · External communityPost link
External question — Cross Validated Stack Exchange
Author: Aaliya Ahamed
Original post: https://stats.stackexchange.com/questions/669861
License: CC BY-SA 4.0 — https://creativecommons.org/licenses/by-sa/4.0/
Adaptation: HTML converted to plain text; contact email addresses removed.
I have a cross-sectional dataset with ~700 children nested within 616 families (siblings share the same
famid
). I want to test associations between exposures and outcomes while accounting for the non-independence of siblings.
I’m currently using generalized estimating equations (GEEs) via
geeglm
in
R
, with
famid
as the clustering variable and an exchangeable working correlation structure.
Would a mixed-effects model with random intercepts for family be more appropriate in this case? Since the data are cross-sectional (one row per child), does GEE essentially just behave like cluster-robust SEs? What are the trade-offs in assumptions and interpretation between GEE and mixed models here?
I am new to statistics, and would like to know thoughts which I could easily be ignoring.
Quote
Report
Rick Hass · External communityPost link
External answer — Cross Validated Stack Exchange
Author: Rick Hass
Original post: https://stats.stackexchange.com/a/669872
License: CC BY-SA 4.0 — https://creativecommons.org/licenses/by-sa/4.0/
Adaptation: HTML converted to plain text; contact email addresses removed.
Though you have cross-sectional and not longitudinal data,
this answer
addresses the key differences between GEE and mixed models, and
this answer
goes into more depth. For a more detailed exposition see
this paper
. As a quick summary, the two approaches yield very similar results for "ordinary" multilevel regression (continuous outcome) as well as a log-link generalized version. If you are using logistic regression, you must be very careful when interpreting the results of the mixed-model.
Your choice also depends on whether you have variables at the family level that you want to interact with individual level variables. For example, say you want to test whether some key exposure effect at the individual level varies across families with different incomes. That is a cross-level interaction, and is more properly dealt with in a mixed-model.
Quote
Report
Post Reply
Quoted from Forex.com.bd-Editorial External answer — Cross Validated Stack Exchange Author: Rick Hass Source score (net votes, not local likes): 3 Original post: https://stats.stackexchange.com/a/669872 License: CC BY-SA 4.0 — https://creativecommons.org/licenses/by-sa/4.0/ Adaptation: HTML converted to plain text; contact email addresses removed. Though you have cross-sectional and not longitudinal data, this answer addresses the key differences between GEE and mixed models, and this answer goes into more depth. For a more detailed exposition see this paper . As a quick summary, the two approaches yield very similar results for "ordinary" multilevel regression (continuous outcome) as well as a log-link generalized version. If you are using logistic regression, you must be very careful when interpreting the results of the mixed-model. Your choice also depends on whether you have variables at the family level that you want to interact with individual level variables. For example, say you want to test whether some key exposure effect at the individual level varies across families with different incomes. That is a cross-level interaction, and is more properly dealt with in a mixed-model.
Checking account access…