Multinomial Logistic Regression

#1
Hi everybody!

I have a hmework I gotta do for my Econometrics course in college. I am testing the Dobson's ( An Introduction to generalized Linear m.) examples of multinomial regresion ( nominal in chapter 8 ).
First I obtained the 300 individual cases of the Car preferences problem , then I run it in Minitab, SPSS and R.

I was surprised when I saw that only Minitab got the results as the book! However neither R or SPSS got them! Actually the last two obtained the same result. The difference is in the interception estimator.

I tried and read a lot of manuals of SPSS in order to get the correct estimation, but nothing so far!

i went like this -->Analyze--> Regresion --> Multinomial logistic
There I put my answer as the grade of importance with reference customized in 1 ( no or less difference) , the ages and sex were my factors.

I don't know if you could help me ....I really need those results!

Thanks in advance!
 

trinker

ggplot2orBust
#2
I was surprised when I saw that only Minitab got the results as the book! However neither R or SPSS got them!
Now I'm wondering... Can you provide the R code you used to analyze the data (including the data set)? That would allow us to look it over and see if you perhaps missed something etc.
 

Dason

Ambassador to the humans
#3
It could also be the way the program codes the dummy variables. I know R uses a "set first to 0" constraint on factors - I don't know how minitab or SPSS handle this but it might explain some discrepancy.
 
#4
Re: Multinomial Logistic Regression
here's my data set ............. https://rapidshare.com/files/1022309057/logistico.txt
I used the variables rpta, sex, ed1 and ed2. My reference level was 1 in rpta.
Perhaps I missed something in SPSS ' options, or wrote a bad R codification. Thanks a lot for the help!

R code:
Using package mlogit
data<-read.table("logistico.txt", header=T)
attach(data)
names(data)
data$rpta<-as.factor(data$rpta)
logdata<-mlogit.data(data,choice="rpta", shape="wide")
modelo<-mlogit(rpta~1|sex+ed1+ed2, data=logdata,reflevel="1")
summary(modelo)