Statalist


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

Re: Re: st: keeping likelihood for every observation


From   Maarten buis <maartenbuis@yahoo.co.uk>
To   Stephane Mahuteau <smahutea@efs.mq.edu.au>
Subject   Re: Re: st: keeping likelihood for every observation
Date   Sun, 7 Feb 2010 04:34:08 -0800 (PST)

--- On Sun, 7/2/10, Stephane Mahuteau wrote:
> Thank you for replying to my question. Just a little
> precision to make sure I get what you wrote. In the line
> "qui replace `lnf'=lnnormalden($ML_y1,`xb'......etc, I need
> to replace everything that is after the equal sign by the
> likelihood of the model I am estimating, isn't it? I am
> comparing two models, a double hurdle model with
> independence and another with dependence and I wanted to
> compute both the Vuong test and the Distribution free test
> (Clarke). For taht I need the value of the likelihood of
> each observation to compute the statistics. So if I
> understand what you say I would write the likelihood for the
> model with independence in the line starting by `lnf' and
> run the program so I get the likelihood for each
> observation, I do the same with the model with dependence
> and then I can compute my stats.
> Is there an easier way for me to do that after I run my
> double hurdle command as a post estimation output? 
> like craggit premit $zlist, second(ihs_remv2_defl $toblist)
> cluster(i)
> and then followed by the appropriate predictnl double lnf=
> ????

It is better to ask such follow up questions on the list, as
somebody else may have also something interesting to say, or
others may be interested in here this discussion.

The line qui replace `lnf' is just there to estimate the model, 
it has nothing to do with your question. The answer to your 
question is in the -predictnl- part of my example.

What is often easy to -predict-, and which is often more 
useful, are the first derivative for each equation, the so 
called scores.

Hope this helps,
Maarten

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

http://www.maartenbuis.nl
--------------------------


      

*
*   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