[R-sig-Epi] function and glm

[Martyn Plummer]

On Thu, 2006-03-02 at 11:34 +0100, Henrik Kallberg wrote:
> Hi all!
> I've a question for which I have found no answer. I want to create a function that contains the glm function as a kind of subfunction within the larger function. I want to be able to  chose how many predictor variables will be included in the glm function based on the "larger" function. Example:
> function(ab#=dataset#,case#response variable#, exp#exposure variable#, confounders#list of confounder variables#){
> Glmf<-glm(case~exp + confounders, family=binomial(link="logit"), data=inter, na.action=na.fail)
> summary(Glmf)
> return(summary(Glmf))
> }
> I want to be able to put in different numbers of confounding variables. I would be very GRATEFUL for suggestions how to solve the problem.
> Best Regards Henrik Kllberg
> henrik.kallberg at ki.se

A dot on the right hand side of the formula means "regress on
everything". So you can create a data frame with only the variables you
want and pass this to glm inside your function

mydata <- cbind("case"=case, "exp"=exp, as.data.frame(confounders))
glm(case ~ ., data=mydata)

Martyn

-----------------------------------------------------------------------
This message and its attachments are strictly confidential. ...{{dropped}}