Linearity of logit for Multinomial logistic regression

ajef

New Member
#1
Hi

I'm trying to test the assumptions for multinomial logistic regression:
How can I test the linearity of logit for my continuous independent variables for more than 2 outcomes? I can only find descriptions for binary logistic regression..


Thanks!
 

hlsmith

Not a robit
#2
I am guessing you could use the ?box tidwell? trick of the including a transformed term in the model. I like fitting a general additive model fitting a spline like term and just looking a the relationship between the continuous predictor and outcome. I have done either of these with a multinomial model but dont know why they wouldnt work.
 

ajef

New Member
#3
I am guessing you could use the ?box tidwell? trick of the including a transformed term in the model. I like fitting a general additive model fitting a spline like term and just looking a the relationship between the continuous predictor and outcome. I have done either of these with a multinomial model but dont know why they wouldnt work.
Thanks a lot, I wasn't sure if I could use the box tidwell approach for multinomial regression, but in that case I'll give it a go :)! Any suggestions on what to do with variables that turn out to not meet the linearity assumption?