Statalist


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

Re: st: RE: AIC xtpcse


From   ghislain dutheil <gdutheil@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: RE: AIC xtpcse
Date   Sat, 30 Aug 2008 11:15:39 +0200

Stas Kolenikov a écrit :
> On 8/27/08, Martin Weiss <martin.weiss@uni-tuebingen.de> wrote:
>   
>> Try
>>
>>  **********
>>  webuse grunfeld, clear
>>     xtset company year, yearly
>>     xtpcse invest mvalue kstoc
>>         eret li
>>     estimates store new
>>  estimates table new, stats(aic) style(oneline)
>>  **********
>>
>>  to see that this is the normal behavior. Likely explanation: -xtpcse- does
>>  not return the information you are asking for. Type -eret li- to see what it
>>  does return.
>>     
>
> Deeper explanation: what's the sample size you want to plug in to
> compute BIC? # of panels or the total # of observations? Or some
> effective # in between? The AIC and BIC are derived as certain
> asymptotic approximations for i.i.d. data. Their generalizations to
> clustered/structured/hierarchical data are not straightforward, so it
> makes perfect sense that Stata DOES NOT report them and thus protects
> you from using unjustified methods.
>
>   

That's to say that i can't have a criteria to choose between my models 
when i use xtpcse... Is there others criteria i can use?

Thanks
Ghislain


*
*   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