Skip to content

GLMing with all possible combinations (automatically)

2 messages · Christofer Bogaso, ilai

#
Dear all, I was working with a GLM model and wondering whether is it
possible to have a model description with all possible main effects as
well as all interactions. Currently I do it manually however this will
become cumbersome if I have large number of variables.

Means suppose I have 3 explanatory variables and want to feed
automatically with all possible combinations like:


glm(Response~1)

glm(Response~X)
glm(Response~Y)
glm(Response~Z)

glm(Response~X+Y)
glm(Response~X+Z)
glm(Response~Z+Y)

glm(Response~X+Y+X*Y)
glm(Response~X+Z+X*Y)

.................................................... all combinations
which are possible

Thanks and regards,
#
Hi
try
glm(Response~ .^2, data=yourdata.frame)

For all predictors (.) and 2-way interactions (^2).
You might also want to see ?drop.terms and ?formula for automating the
construction of all model combinations.

Side note: R is not SAS (fortunately). Interaction is denoted ":", X*Y
is shorthand for X+Y+X:Y. Your "X+Y+X*Y" is redundant.

HTH

On Wed, Mar 7, 2012 at 4:16 AM, Christofer Bogaso
<bogaso.christofer at gmail.com> wrote: