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

Re: st: Negative LR test statistic ?

From   Maarten buis <>
Subject   Re: st: Negative LR test statistic ?
Date   Tue, 22 Dec 2009 00:31:44 -0800 (PST)

--- On Tue, 22/12/09, Ekrem Kalkan wrote:
> in my case  every equation has 860 number of
> observartion. 

I interpret that as meaning that your dataset contains
860 observations, and you are estimating a model with
14 equations, is that correct? In that case you have 
only 860/14= 61.4 observations per equation, which was 
my point.

> Still, I think there should be a different
> reason rather than degrees of freedom. 

I think degrees of freedom is the reason why you get 
negative likelihood ratio statistics. Remember that 
the reasoning behind the likelihood ratio test is 
asymptotic, so strictly speaking it is only valid in 
case of a infinitly large sample. What sample size is
close enough to infinity depends on the kind of 
information you are trying to extract from: the more 
complex the information you are trying to extract the
larger the sample size needs to be.

Hope this helps,

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


*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index