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: RE: error message: initial values not feasible

From   Maarten buis <>
Subject   RE: st: RE: error message: initial values not feasible
Date   Wed, 7 Apr 2010 08:58:28 -0700 (PDT)

--- On Wed, 7/4/10, sun samn wrote:
>   It is about my simulation: I want to repeat the GMM for
> 1000 times and see how the GMM estimates are under different
> data generation process(DGP). Under certain DGP, the code
> works fine; while when I change the DGP, I get that error
> message. I even tried to change the initial value of the
> parameter, but it did not work either.
>   I guess the method is fine since it works for some DGP;
> while the data or the initial value might not be good. But I
> cannot change the data, you know, I need to stick to that
> setting; so it seems like the only thing I can do is to put
> an appropriate number, right?

Imagine you are someone who tries to answer your question, 
how would the information you profided help in diagnosing the
problem? It is obviously still not enough. I will give 2
guesses at what might be going on, but I will not give much 
details as they may very well not be appropriate to your 

The fact that your evaluator worked on some datasets does
not mean it is good: some parameterizations are more 
numerically stable than others. See: William Gould (2006)
Mata Matters: Precision. The Stata Journal, 6(4):550-660.

Your simulated datasets may just be too extreme: e.g. 
very little variation on some variables or very high 
correlations. Look at the datasets that are causing 
problems and make a decision whether they are worth 
going through much trouble to estimate a model in it. 

-- Maarten

Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen


*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index