I have made a multinomial model using multinom() from the nnet package. I am trying to do internal validation and update the model but I need the linear predictor values from the model and I cannot seem to access them. I can get the predicted probabilities but cannot find a way to get the linear predictor terms.
I have tried predict(MultiNom1, type = "link") but it seems that you cannot use "link" as an argument for the multinomial mode. Wondering if anyone had any suggestions?
e.g.
require(nnet)
require(rms)
attach(mtcars)
car <- multinom(gear ~ am + cyl + rcs(mpg, 4))
predict(car, type = "link")