[R] Multinomial logistic regression with non-numerical data.

[TL Vincent]

Hi all,

I have a set of data which looks similar to this (only way bigger!)

link    type    register

F         D         a
E         D         b
R         D         c
A         T         d
D         T         a
F         D         b
.         .         .
.         .         .

and would like to perform a multinomial logistic regression on this data.
I use the nnet package and obtain the raw regression coefficients from the 
output when I enter:

fit<-multinom(link~type+register,data)
summary(fit, Wald =T);


However, I am struggling to compute the regression results in terms of 
probabilities, my main barrier being the fact that my data is non-numerical.
For example, I would like to fix a value "d" to register, and calculate the 
probability of choosing each link for both type "D" or "T" .
I hope this was clear enough!
I appreciate your help!


Tom V.


----------------------
TL Vincent
tv3190 at bristol.ac.uk
BCCS - Bristol Centre for Complexity Sciences