Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

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

Re: st: question about 3sls using the reg3 command

From   John Antonakis <>
Subject   Re: st: question about 3sls using the reg3 command
Date   Fri, 21 May 2010 09:14:23 +0200


As Austin [hesitantly] suggested, you should use -cmp-, and this, particularly because it is clear that you have ordered probits as dependent variables (so, whatever the limitations are for using OLS with ordered categorical data apply here).

The one limitation with -cmp- is that it does not provide tests of overidentification. Mplus is a program that could correctly estimate this model.



Prof. John Antonakis, Associate Dean Faculty of Business and Economics
Department of Organizational Behavior
University of Lausanne
Internef #618
CH-1015 Lausanne-Dorigny

Tel ++41 (0)21 692-3438
Fax ++41 (0)21 692-3305

Faculty page:

Personal page:

On 21.05.2010 03:13, JOSE A ALEMAN wrote:
Hi Austin,

To answer your question, my dependent variables can only take the values 0,
1, 2, 3, 4, 5, 6, 7, or 8 in one case, and 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 or
10 in the other. How would this affect estimation?



Austin Nichols <austinnichols@gm> To Sent by: owner-statalist@h cc sphsun2.harvard.e du Subject Re: st: question about 3sls using the reg3 command 05/20/2010 08:55 PM Please respond to statalist@hsphsun

In short, no, but see

-reg3- is for linear models, like OLS or -ivreg2-, and ordinal or
fractional or nonnegative dependent variables present problems; there
are several alternative approaches which are not very appealing, but
-cmp- is probably your bet best.  You should probably compare your
output there to the closest linear analog of each model using -ivreg2-
since -ivreg2- will give you a wealth of specification tests and
simple interpretations (all of which will be wrong, but might be
approximately right, given the hypothesized nonlinearity).

Do your dependent variables take on any real value between 0 and N, or
integers or some other small subset, and are the endpoints included?
l.e. is y1 in [0,8] or (0,8) or {0,1,...8} or what?  These are
important distinctions...

On Thu, May 20, 2010 at 6:52 PM, JOSE A ALEMAN <> wrote:
Dear all,

I would like to estimate a system of simultaneous equations where my
dependent variables are both intervals (one from 0 to 8 and the other
0 to 10). Each dependent variable is endogenous to the other in the
of equations. My question is if the stata command reg3 works with such
nonlinearities? If so, is there a reference text/document that explains
this model and provides formulas that depict how estimation is achieved?

Thank you,

Jose A. Aleman, Ph.D.
Assistant Professor
Political Science Department
Fordham University
*   For searches and help try:

*   For searches and help try:
*   For searches and help try:

© Copyright 1996–2016 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index