Going from a list of variable names to a valid formula call

Question

I have dataset df

df<-data.frame(outcome=sample(c(0,1),100,T),
               v1=rnorm(100,10,2),
               v2=rnorm(100,10,2),
               v3=rnorm(100,10,2),
               v4=rnorm(100,10,2),
               v5=rnorm(100,10,2))

And I make a list of variables I want to look at in a later model:

list<-c("v1","v3","v5")

And I want to be able to manipulate list in order to enter it into a glm model:

magical steps
.
.
.

To give the desired formula call:

glm(outcome~v1+v3+v5,data=df,family="binomial")

Where glm(outcome~ and ,data=df,family="binomial") are preset, but v1+v3+v5 is the result of manipulations on list.

When my list is only a length of 1, I can get away with ~ df[,list], but when length(list) > 1 I get:

ERROR: R: invalid type (list) for variable 'df[, list]'

I've also tried creating a smaller dataset:

predictors<-df[,list]

and entering that in:

glm(outcome~predictors,data=df,family="binomial")

but I get a similar error:

ERROR: R: invalid type (list) for variable 'predictors'

Thank you!

see: http://stackoverflow.com/questions/29555473/creating-formula-using-very-long-strings-in-r — Zelazny7, Jun 24 '15 at 14:19
also see [this](http://stackoverflow.com/q/12967797/324364). — joran, Jun 24 '15 at 14:23

Going from a list of variable names to a valid formula call

0 Answers0