Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Likelihood-ratio test for frontier models


From   "Eva Poen" <eva.poen@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Likelihood-ratio test for frontier models
Date   Tue, 16 Sep 2008 15:01:04 +0100

Monika,

with regard to your first question:
A missing for ll(null) in the output of -lrtest- is nothing to worry
about (and if you look back in your output, they should be missing in
your estimations already). The null likelihood refers to the
likelihood of a model where the only explanatory variable is a
constant. If you did not include a constant term in your models
(option -nocons-), Stata will report a missing value for e(ll_0) in
the regression output, and also in the output of -lrtest-.

This doesn't interfer with the actual likelihood ratio test that
-lrtest- performs (it compares the likelihood of a restricted and an
unrestricted model). Martin already commented on how to interpret the
test result.

Eva


2008/9/16 Monika Baron <baronmonika@googlemail.com>:
> Hello everyone!
>
> I'm modelling a Stochastic Cost Frontier with the command frontier.
> I'm trying to find the model's best functional form by using a
> Likelihood-ratio test.
>
> My output:
> ********
> . lrtest A B, stats
>
> Likelihood-ratio test                                  LR chi2(6)  =     49.42
> (Assumption: B nested in A)                            Prob > chi2 =    0.0000
>
> -----------------------------------------------------------------------------
>      Model |    Obs    ll(null)   ll(model)     df          AIC         BIC
> -------------+---------------------------------------------------------------
>          B |    439           .    548.5923     17    -1063.185    -993.748
>          A |    439           .    573.3007     23    -1100.601   -1006.658
> -----------------------------------------------------------------------------
>              Note:  N=Obs used in calculating BIC; see [R] BIC note
>
> *********
>
> 1. Why are there no ll(null) in the statistics?
>
> 2. What do the results imply? Which model should I choose?
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index