On Nov 16, 2009, at 2:31 PM, shuai yuan wrote:

Hi, there,

My appologize if someone ask the same question before. I searched the
mailing list and found one similar post, but not what i want.

The problem for me is, I use the step( glm()....) to do naive forward
selection for logistic regression.  My code is functional
in the open environment. But if I wrap it up as a function, then R keeps
saying "object 'a' not found". Actually,  data frame
"a" is inside the function.

I did some search online. i guess the reason may be R did not keep the data
in glm() output after building the model but not sure.
Can anyone please tell me how to work around this problem?

Thanks a lot in advance.

I am using R 2.9.0. Here is the sample code:

#------------
naivelr<-function(x,y){
:
:
:

a<-data.frame(x)

form<- paste("y~1+",paste(grep("X.*",names(a),value=T),collapse="+"),sep="")

if(is.null(force.in)!=T){
lowmo<-paste("y~1+",paste(grep("X.*",names(a) [force.in],value=T),collapse="+"),sep="")

} else
{lowmo<-"y~1"}

lower1<-glm(lowmo,family="binomial",data=data.frame(a,y))
upper1<-glm(form,family="binomial",data=data.frame(a,y))

You are sticking data.frame= a inside another data.frame ????



stepout<- step (lower1 ,scope =list(lower=lower1,upper=upper1),direction="forward",k=0,trace=100)

Thats not the way I remember "step"-ping. I thought you made a fit and then "stepped" the formulas (using the same data), rather putting the whole glm object into a "lower" and an "upper". I could be wrong about that since I try to avoid using stepwise methods.

# here is the error:Start:
#AIC=689.62
#y ~ 1
#Error in data.frame(a, y) : object 'a' not found---but "a" is there!

But it's probably not in a form that can be interpreted. Consider adding y as a column in "a".

David Winsemius, MD
Heritage Laboratories
West Hartford, CT

______________________________________________
R-help@r-project.org 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.

Reply via email to