question about MuMIn package

Dear R users,

I would like to use the MuMIn package to perform model average on a
simple GLM containing a mixture of categorical and continuous
predictors. The model.avg functions seems to work with models
containing continuous and categorical variables, but I am wondering
whether somebody could tell me if the relative importance obtained for
categorical predictors (i.e. the selection probability) is meaningful
when using this function?

thanks in advance,

Simone