Skip to content

empty cell when running GEE for binary data

3 messages · Özgür Asar, Yue

Yue
#
Hi all,

My data are binary and within-subject correlated; there are three factors of
interest and two of them are within-subject. So, I have considered modelling
the data using a binomial regression with a GEE approach (which can be
achieved using the function geeglm). One problem is that I got some empty
cells, so I can't simply use logit as a link function. I was wondering if
you know any existing R solution for this? 

also, I tired to use "cauchit" as the link function, but it turned out that
within the function geeglm "cauchit" is not available. Any idea on this? 

Many thanks!
Yue

--
View this message in context: http://r.789695.n4.nabble.com/empty-cell-when-running-GEE-for-binary-data-tp4635483.html
Sent from the R help mailing list archive at Nabble.com.
Yue
#
hi Ozgur,

Thanks for your reply! by "empty cell" I mean there was a condition in which
all participants showed the same response (let's say response A). In other
words, the probability of A, according to the present data, was 100%; this
led to the empty cell for non-A. The odds or p(A) would be 1/0, so it's not
appropriate to use logit as the link function. 

I don't think, but not sure, gee would work, but i will try the other
package you suggested! 

Thanks!
Yue

--
View this message in context: http://r.789695.n4.nabble.com/empty-cell-when-running-GEE-for-binary-data-tp4635483p4635513.html
Sent from the R help mailing list archive at Nabble.com.