R help - Jul 2011 - Missing values and geeglm

Dear all

I am struggling with how to deal with missing values using geeglm. I know
that geeglm only works with complete datasets, but I cannot seem to get the
na.omit function to work.  For example

assuming DataMiss contains 3 columns, each of which has missing
observations, and an id column with no missing info then identifies the
clusters.

Outcome: 2 level integer
Predictor: numeric variable
Confounder: 3 level integer

 If I "manually" remove the missing values then run the model, there
is no
problem. 

#remove missing values
data<-subset(DataMiss, !is.na(outcome) & !is.na(predictor) &
!is.na(confounder))

#run the model
model<-geeglm(outcome~predictor+confounder, family=binomial(link =
"logit"),
data=data, corstr='ar1', id=id, std.err="san.se")

However, I don't always want to have to run this extra step. The R
instructions seem to indicate that na.omit should work, as shown below

model<-geeglm(outcome~predictor+confounder, family=binomial(link =
"logit"),
data=na.omit(DataMiss), corstr='ar1', id=id, std.err="san.se")

But I keep getting this error. Any help would be greatly appreciated!

Error in family$linkfun(mustart) : 
  Argument mu must be a nonempty numeric vector







--
View this message in context:
http://r.789695.n4.nabble.com/Missing-values-and-geeglm-tp3675622p3675622.html
Sent from the R help mailing list archive at Nabble.com.

model<-geeglm(outcome~predictor+confounder, family=binomial(link =
"logit"),
data=na.omit(DataMiss), corstr='ar1', id=id, std.err="san.se")

There could be other variables in DataMiss that have many missing values, so
when you apply na.omit() on DataMiss, you may be ending up with an empty
data.frame. Try subsetting DataMiss to only the variables you will use for
the model, and then taking na.omit of it outside of your model statement.

DataMissOm <- na.omit(subset(DataMiss, select = c(outcome, predictor,
confounder, id)))
model<-geeglm(outcome~predictor+confounder, family=binomial(link =
"logit"),
data=DataMissOm, corstr='ar1', id=id, std.err="san.se") 



--
View this message in context:
http://r.789695.n4.nabble.com/Missing-values-and-geeglm-tp3675622p4649433.html
Sent from the R help mailing list archive at Nabble.com.