Dear R users,
I've got a serious problem running some gee functions, and I really can't
fix it. My dataframe is made of several rows and columns (say 7600 x 15),
like the one below:
header inr ......... inside group_cod
1 2.25 0 1
1 3.46 0 0
1 ...... 1 0
1 ...... 1 .......
1 ...... ..... .......
.... ...... ..... .......
.... ...... ..... .......
.... ...... ..... .......
.... ...... ..... .......
.... ...... ..... .......
615 ...... ..... .......
615 ...... ..... .......
615 ...... ..... .......
As you can see I've got several repeated measures, resulting in clusters (
i.e. ID) of different sizes. You can get a frame like this by running this
code:
header <- rep.int(seq(1:615),sample(seq(1:19),size=615,replace=T))
inr <- rlnorm(length(header),0.8434359,0.3268392)
group_cod <- sample(c(0,1),size=length(header),replace=T)
inside <- sample(c(0,1),size=length(header),replace=T)
gee.frame <- data.frame(header,inr,group_cod,inside)
When I try running a longitudinal model with geeglm, the
corstr="unstructured" option returns me an assertion failure of R itself (a
sort of error window), while the ar1 option for corstr returns a model only
with estimates, and no s.e. or wald test. Same is for the anova method. I
removed subjects with only one observation, but the problem's still the
same. This is the code used:
- for corstr= unstructured
geeglm.model<- geeglm(inside~group_cod,family=binomial, data=gee.frame
,id=header,corstr="unstructured")
This model gives me an error (not the kind written in R-gui: simply the
program seems to stop not producing anything), and doesn't produce any
result, leading me to quit R. The same happens when I use a userdefined
matrix built with genZcor function using crostrv=4 (i.e. unstructured). My
initial thought was of problems because of highly unbalanced observations
and because of exceeding of correlation parameters to estimate, but
restricting cluster size did not result in any improvement.
Is there a problem with the code, or it could be due to the data?
- for costr= ar1
In my dataset I also have a variable "weeks", that I use to specify waves in
the geeglm. When I use this, the output for an autoregressive gee model
gives me only the estimates, when I don't use it everything's alright. The
problem's that I have unbalanced observations, so the use of waves could
take this into account while estimating parameters. Is this a conflict
between these options, or what?
In addition, I've built the empirical correlation structure. A 19x19
structure, thus with nrow=maximum cluster size. When I use this as zcor the
program gives me
geeglm.fixed <- geeglm(inside~weeks+group_cod+età,family=binomial, data=
frame.model,id=header,waves=weeks,zcor=corr.gee,corstr="userdefined")
*Errore in geese.fit(xx, yy, id, offset, soffset, w, waves = waves, zsca, :
nrow(zcor) need to be equal sum(clusz * (clusz - 1) / 2) for
unstructured or userdefined corstr.*
I really don't understand the meaning of this: the correlation should have
number of rows equal to maximum cluster size. Also in the online help the
details say something about the dimension of Zcor, but I can't understand
the same...
I hope I've been clear.
Thanks in advance
niccolò
[[alternative HTML version deleted]]
______________________________________________
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
