Re: st: IIA test for multinomial logit

[Richard Williams <richardwilliams.ndu@gmail.com>]

At 09:52 AM 1/29/2014, Stepien, Paulina wrote:
> Dear Stata Users,
>
> I am trying to perform an IIA test for multinomial logit where 
> dependent variable employment status (empl_st6) has 4 categories: 0 
> unemployed, 1 employed for wages, 2 self-employed, 3 employer. I 
> have tried running the test in two ways, none of which seem to be working.
>
> Firstly, I have run mlogit regression followed by mlogtest, smhsiao. 
> This command returns following error messages: 'problem determining 
> number of categories' and 'factor variables and time series 
> operators not allowed'.


-findit mlogtest- yields several versions of the program. The oldest 
dates way way way back to the Stata Technical Bulletin. I believe the 
last official release was in spost9_ado. If you are really really 
brave you can get the beta versions of Long and Freese's new spost 
programs contained in test13_ado. Personally, I would try the spost9 
version if that isn't what you already have. Usually, when there are 
multiple versions of a program I assume that the SSC version is best, 
but that isn't true for Long and Freese programs because they 
originally released programs on STB or SSC but then switched to 
keeping all the latest and greatest versions on Long's own web site.

The error "'factor variables and time series operators not allowed'" 
probably means exactly what it says. If you did something like

mlogit y i.x1 x2
mlogtest

your version of mlogtest may not support the use of factor variables. 
You would therefore have to create the dummy variables yourself. Many 
older user-written programs (and some newer ones) won't support 
relatively new features of Stata like factor variables.

In short, if you stop using factor variables, or get a newer version 
of mlogtest, your problems may go away. But then again maybe not. 
Clean these things up and then write back if problems persist. There 
is no point in trying to figure out the problems until you have tried 
the easy fixes first.



> I have looked at the dependent variable specification but not sure 
> what may be wrong with it. I have also looked at all potential 
> factor variables and time series markers but even after eliminating 
> potential culprits the test does not work. One advice I have found 
> on stata blog is to continue setting seeds until it works, which I 
> have tried with no success. I am not sure how many times I should 
> try before I give up.
>
> Secondly, I have tried running smhsiao empl_st6, elim(0) but two 
> problems occur here: the regression fails to calculate chi2 and an 
> error message comes up saying 'equation unemployed not found'. When 
> I recode categories to run from 1-4 instead 0-3, the error message 
> does not come up but chi2 is still not being calculated.
>
> I would be most grateful if someone could advise what I am doing 
> wrong or whether there is any solution to this problem.
>
> Thank you in advance.
>
> Paulina
>
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/

-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
HOME:   (574)289-5227
EMAIL:  Richard.A.Williams.5@ND.Edu
WWW:    http://www.nd.edu/~rwilliam

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/