Statalist


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

Re: Re: st: double clustering


From   "Austin Nichols" <austinnichols@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: Re: st: double clustering
Date   Sun, 8 Jun 2008 21:19:33 -0400

You can treat country as the primary sampling unit in your -svyset-
command, even though it actually includes many primary sampling units,
assuming you have many countries in your sample.  This allows for
arbitrary intracountry correlation of errors, and the lower-level
(school) PSU clustering is essentially subsumed (i.e. the change to
estimated SEs from specifying two levels of clustering in -svyset-
will be small, and the justification more problematic).

On Tue, Jun 3, 2008 at 1:19 PM,  <nicola.baldini2@unibo.it> wrote:
> Back on the double clustering: the command is -creg-, which I should have downloaded somewhere around from http://www.kellogg.northwestern.edu/faculty/petersen/htm/papers/se
> However, I am not sure it is panel command. Have a look at http://www.nber.org/papers/T0327 too
>
> Nicola
> At 02.33 01/06/2008 -0400, Jeph Herrin wrote:
>>See -xtmixed- and -xtmelogit-.
>>
>>Orsetta.CAUSA@oecd.org wrote:
>>> Hello
>>>
>>>
>>> I am running country-by-country regressions on a sample of students, that are
>>> clustered in schools. I cluster at the school level. Now, if I want to pool
>>> the countries, I still need to take school cluster into account, but I would
>>> like to also cluster at the country level. I know that stata allows double
>>> stage sampling in svy, but I don't think it is correct to consider the
>>> country as the psu (and the school would be the ssu), because a country is
>>> not a sampling unit since it is sampled with certainty.
>>> I remember having noticed in some old statalist archive that there is a
>>> user-written double clustering option available (ideally one would want to
>>> explicit the robust cluster option followed by 2 variables). does anybody has
>>> a clue on it? What would you suggest to do? (not considering gllam modeling
>>> since my programs are far too long and heavy and would not run in a gllam
>>> configuration).
>>>
>>> Many thanks :)
>>>
>>> orsetta
*
*   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/



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