The issue of endogeneity is different for logistic models. If your
data is aggregated, i.e. a multiway table, you could estimate a
multinomial logistic model as a loglinear model using -glm- or
-poisson-. In loglinear models, there's no distinction between
dependent and independent variables, there's no causal direction. All
you're interested in is whether there's a relationship among the
variables.
So one course you can take is to recast your model as a loglinear
model. Or you could note the equivalence between multinomial logit
and loglinear models to show that endogeneity is not an issue.
Final note: unfortunately, you can't solve the "polychotomous"
problem by referring to the models as "multinomial". I've seen this
usage criticized on other lists too :-(.
Good luck,
John Hendrickx
--- "Ngo,PT (pgr)" <[email protected]> wrote:
> Dear Statalisters,
>
> Sorry to bother you again for the second time in the day!
>
> I would like estimate a polychotomous logistic model using mlogit.
> The main explanatory variable (say X) I use is endogenous and in
> binary models, I have used IVs using the ivprob command. How would
> one go about estimating polychotomous logistic model with an
> endogenous variable for which I have an instrument? I am interested
> in getting the relative risk ratio as I am trying to differentiate
> the impact X on the various discrete categories of Y on the left
> hand side.
>
> Any advice would be really appreciated. It is the first time I am
> using logit/probit and
> polychotomous models.
>
> Thanks a lot in advance,
>
> Thi Minh
>
>
>
>
> *
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
_______________________________
Do you Yahoo!?
Declare Yourself - Register online to vote today!
http://vote.yahoo.com
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
