# Re: st: SUR and a system of logit models

 From "Michael Anthony Carlton" <[email protected]> To [email protected] Subject Re: st: SUR and a system of logit models Date Sat, 26 Apr 2008 16:48:05 -0400

You are correct. Estimating these equations as a system will only gain efficiency. Because the logit is in the linear exponential family as long as the conditional mean function is modeled correctly you will have consistent parameter estimates.
F. Gao writes:

Dear all,
I actually have more than 2 equations that I am currently running
logistic regressions separately.
My understanding is that estimating them together would increase the
efficiency of the estimates. But is there anything strongly against
doing this instead of estimating a system of equations?
Any other suggestions?
Thanks!
Faye
If there is no STATA procedue, is there SAS procedure?
On 4/26/08, Phil Schumm <[email protected]> wrote:

On Apr 25, 2008, at 4:53 PM, F. Gao wrote:
> What to do if I have a system of logit models (binary dependent variables)
and wish to do SUR regression?
> eg: Pr(A=1)=f(x,y), Pr(B=1)=f(x,z)
>
On Apr 26, 2008, at 5:35 AM, Stephen P. Jenkins wrote:
> There is no such approach that I am aware of, as there is no multivariate
logistic distribution.
>

One possible approach is alternating logistic regression:
V. Carey, S. L. Zeger, and P. Diggle. Modelling multivariate binary data
with alternating logistic regressions. Biometrika, 80(3):517–526, 1993.
I am not aware of a Stata implementation.

-- Phil

*
* 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/
*
* 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/
```

*
*   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/
```