6
$\begingroup$

I have a question regarding the interpretation of the output I get by fitting a GEE model in R. Attached is the picture of the output I get: enter image description here

1) The first column, Estimate, is obviously the estimated $\beta$'s under the GEE model.

3) The third column seems to be the Z statistic calculated using the Naive S.E

5) Fifth column seems to be the Z statistic calculated using the Robust S.E

2) Naive S.E: I guess this one is the estimated S.E using the correlation structure I used in my GEE (here: corstr = "exchangeable")

4) Robust S.E: I guess is the S.E estimated using the empirical sandwich estimator.

Is that right?

$\endgroup$
1
$\begingroup$

Naive SE is the same SE you would get if you perform GLM

Robust SE is SE calculated using robust sandwich estimator.

This vignettes explained GEE pretty well with an example https://cran.r-project.org/web/packages/HSAUR2/vignettes/Ch_analysing_longitudinal_dataII.pdf

Hope that helps!

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ "Naive SE is the same SE you would get if you perform GLM" is not true. For example, type this: x = rnorm(50); id = rep(1:10, each=5); y = rnorm(50)>0; m1 = gee(y ~ x, id=id, family=binomial); m2 = glm(y ~ x, family=binomial ) $\endgroup$ – Do not reinstate Monica Jul 2 at 17:31
  • $\begingroup$ It would be true if "independence" is assumed and the scale parameter is accounted for. So, the naive SE given by m3 = gee(y ~ x, id=id, family=binomial, scale.fix = 1) is same as m2. Alternatively, from m1, dividing the naive SE by sqrt(Estimated Scale Parameter) gives the SE of m2. $\endgroup$ – T Lin Aug 14 at 18:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.