Re: st: Test Rho

[Garry Anderson <g.anderson@unimelb.edu.au>]

Hello Irene,

My interpretation is that it is testing if the panel-level variance 
component (sigma_v^2) is zero.

rho = sigma_v^2 / (sigma_v^2 + sigma_e^2)

which is the proportion of the total residual variance contributed by the 
panel-level variance component.

However, in my opinion, it is the variance inflation factor (VIF) that is 
more important.

VIF = 1 + (n' - 1) rho

where n' is the average number of observations per panel.
If one has a large n', the VIF inflation factor can be substantially more 
than one, even if rho is small. So, a small rho that is not statistically 
significant can have a large numerical VIF.
I suppose it comes down to what is a large VIF, rather than the statistical 
significance of rho?
Is say a VIF of 1.3 large?

It is important that you are using the 12 Jan 2007 update to Stata 9.2 
because versions prior to this could ignore large panels when fitting the 
model. This could effect the value of rho and hence the VIF.

Regards, Garry Anderson
Dept Veterinary Science
University of Melbourne
Australia
g.anderson@unimelb.edu.au


At 03:27 PM 25/01/2007 +0200, you wrote:

> Hello,
>
> From the following test, occuring after an xtlogit with random-effects 
> what is the conclusion for frailty. Is it signicant?
>
> Could someone explain the interpretation?
>
> Likelihood-ratio test of rho=0: chibar2(01) = 1.77 Prob >= chibar2 = 0.092
>
> Thank you very much
>
> Irene
>
>
> *
> *   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/