[S] GEE problems

Hanno Ulmer (Hanno.Ulmer@uibk.ac.at)
Thu, 16 Apr 1998 13:55:06 +0100


Dear S+ users!

I am trying to fit GEE models like the following to my
data (risk factors for headache, cluster sampling) . I always
get these warnings when I use a binomial family, in the case of
poisson I get the estimates. I would be glad if someone could help
me with my problem. (I am using S+ 4.* for Windows95 and GEE 4.13)

best wishes

Hanno Ulmer

head1.gee<-gee(as.numeric(HEAD) ~ SEX+GRD2+DNS2+LIVE+CONT,
+ id=CLUSTER,data=amas97o1,na.action=na.omit,family=binomial)
[1] "Beginning Cgee S-function, @(#) geeformula.q 4.13 9
8/01/27"
[1] "running glm to get initial regression estimate"
[1] 2.34396021 0.14889552 0.01327743 -0.05680151
[5] -0.14591835 -0.07828458 -0.07319605
Cgee: error: logistic model for probability has fitted v
alue very close to 1.
Cgee: estimates diverging; iteration terminated.
Cgee: returning to S
There were 7 warnings (use warnings() to see them)
> warnings()
Warning messages --
1: NAs generated in: log(1 - y[nz])
2: NAs generated in: log(1 - y[nz])
3: NAs generated in: log(1 - mu)
4: fitted values close to 0 or 1
5: NAs generated in: log(1 - y[nz])
6: iterations terminated prematurely because of
singularities in: glm.fitter(x = X, y = Y, w =
w, start = start, offset ...
7: Note: Cgee had an error (code= 1 ). Results
suspect. in: gee(as.numeric(HEAD) ~ SEX + GRD2 +
DNS2 + LIVE + CONT, ...
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news