[R] IIA test on multinomial logit

[ronggui]

Hi, you should look at hmftest in mlogit package. However, be cautious
when you use IIA test, it is not very reliable as Long and Freese
(2006:243) have shown.

* J. Scott Long and Jeremy Freese. 2006. Regression Models for
Categorical Outcomes Using Stata. 2nd.

On Wed, Oct 15, 2008 at 3:37 AM, Erika Known <e at it.dk> wrote:
>> ----- Original Message -----
>> From: "Erika Known" <e at it.dk>
>> To: <r-help at r-project.org>
>> Sent: Tuesday, October 14, 2008 11:43 AM
>> Subject: [R] IAA test on my multinomial logit
>>
>> > Hi all
>> >
>> > I'm trying to immigrate from Stata and I want to preform a IIA test
>> > (Independence of Irrelevant Alternatives) on my multinomial logit model
>> > in R, but I can't seem to find a package that dos the trick. I've been
>> > looking at
>> >
> http://bm2.genes.nig.ac.jp/RGM2/R_current/library/systemfit/man/hausman.systemfit.html
>> > but I can't figure out what to do! I normally just type 'mlogtest,
>> > hausman base' to get the test in Stata. I can even use 'mlogtest,
>> > smhsiao base' to get the Small-Hsiao tests of IIA assumption.
>> >
>> > according to the literature there is also other tests.. thou I cant seem
>> > to find any off them in R.
>> >
>> > Thank for the help.
>> >
>> > Erika
>
> On Tue, 14 Oct 2008 12:11:45 -0500, Gilbert Brenes wrote
>> If I'm not wrong, the test for IIA is the Hausman test.
>>
>> Look for hausman.systemfit in the systemfit package.
>>
>> Gilbert
>>
> Hi again!
>
> Thanks for your answer! Thou that's where I got stuck too, so to say.
> The link I send was to the
> systemfit package, but I can't seem to get it to working.
>
> This is what I do:
>
> model_01 <- multinom(Y ~ X1 + X2 + X3 + X4 + X5 + X6, data=data,
> contrasts = NULL, Hess = TRUE, summ = 0, censored = FALSE, model = TRUE)
>
> That's running like a dream! - and the I try:
>
>> systemfit(model_01)
> Error in systemfit(dyls.model01) :
>  argument 'formula' must be an object of class 'formula' or a list of
> objects of class 'formula'
> In addition: Warning messages:
> 1: In if (class(formula) == "formula") { :
>  the condition has length > 1 and only the first element will be used
> 2: In if (class(formula) == "list") { :
>  the condition has length > 1 and only the first element will be used
>
>>hausman.systemfit(model_01)
> Error in coef(results3sls) :
>  argument "results3sls" is missing, with no default
>
> I relly want to imigrate, but I need help! Just tell mer a website or
> wher I can read my way out of it.
>
> Do you think you can give me one more hint?
>
> Sorry for my ignorance
>
> Erika
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>



-- 
HUANG Ronggui, Wincent
Tel: (00852) 3442 3832
Ph.D. Candidate, CityU of HK
Master of sociology, Fudan University, China
Bachelor of Social Work, Fudan University, China
Personal homepage: http://ronggui.huang.googlepages.com/