[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: Testing IIA assumption with micombine mlogit command

From   Julie Trivitt <>
Subject   RE: st: Testing IIA assumption with micombine mlogit command
Date   Tue, 09 Jun 2009 15:12:37 -0500

Great idea, Richard.  Thanks!

I'll try it on each imputation and the full dataset and hope for consistent

-----Original Message-----
[] On Behalf Of Richard Williams
Sent: Tuesday, June 09, 2009 2:46 PM
Subject: Re: st: Testing IIA assumption with micombine mlogit command

At 03:07 PM 6/9/2009, Julie Trivitt wrote:
>I need to test the IIA assumption with a multinomial logit model.  I 
>have used the mlogtest,all command, but get the following error message:
>Data has been altered since mlogit was estimated.
>I suspect it is since I had to use the ice command to impute missing 
>data and the actual estimation command is micombine mlogit........
>Can anyone advise me on the best way to test the IIA assumption in this
>Thanks in advance,

In cases like this, I wonder how horrible it would be to, say, test 
the IIA assumption on each imputed data set separately.  Sure, the 
numbers wouldn't be exactly right, but it is, after all, a diagnostic 
test.  If you are lucky the data sets will all clearly pass the test 
or all fail it.  If somebody knows how to do it exactly right, then 
great, but barring that i don't think it would be too horrible to 
improvise a bit.  There are a lot of things still being ironed out 
with multiple imputation, so reasonable improvisations may be better 
than doing nothing at all.

Of course, the mlogtest code (part of Long and Freese's spost 
package) is there for everyone to see, so someone might be able to 
adapt it for the multiple imputation situation.

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

*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2022 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index