Statalist The Stata Listserver


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

Re: st: TSCS: what model for N>T ??


From   "joel miller" <joelstar90@hotmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: TSCS: what model for N>T ??
Date   Fri, 10 Mar 2006 09:26:02 +0000

Thanks Clive,

That´s the kind of response I was hoping to get!

Best,

Joel Miller
University of Málaga


From: "Clive Nicholas" <Clive.Nicholas@newcastle.ac.uk>
Reply-To: statalist@hsphsun2.harvard.edu
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: TSCS: what model for N>T ??
Date: Fri, 10 Mar 2006 00:43:37 -0000 (GMT)

Joel Miller wrote:

> I´ve been working on TSCS data involving 41 police force areas, and about
> 7
> time periods (soon I hope to be 8). I´m theoretically interested in a
> fixed
> effects model, because I want to see how variation within police forces is
> impacting arrest rates over time (and Hausman tests suggest a fixed
> effects
> model is appropriate).
>
> However, is OLS fixed effects my only choice here? I have played a little
> with a panel adjusted standard error (the Beck and Katz model). But while
> I
> have seen published articles which use the PCSE model with a similar data
> structure to mine (N a lot greater than T), I have since been advised that
> this is not appropriate. This puzzles me a little, because I have also
> read
> further that the Beck/Katz is actually more suited to N>T than other
> approaches.
>
> In short I´m confused. Help!

Directly from Beck (2001: 274):

"...TSCS methods are justified by asymptotics in T and typically require a
reasonably large T to be useful. ... (T)here is no hard and fast minimum T
for TSCS methods for work, but one ought to be suspicious of TSCS methods
used for, say, T < 10. On the other hand, TSCS methods do not require a
large a large N, although a large N is typically not harmful. ... (P)anel
methods are designed for and work with very small Ts (three, perhaps even
two) but require a large N for the theoretical properties of the
estimators to have any practical consequences."

Since this comes from the doyen of applying TSCS models to
political/social data, my recommendation would be to run -xtpcse- models
on your data anyway, but present and interpret them with huge caveats.

CLIVE NICHOLAS |t: 0(044)7903 397793
Politics |e: clive.nicholas@ncl.ac.uk
Newcastle University |http://www.ncl.ac.uk/geps

Reference:

Beck N (2001) "Time-Series Cross Section Data: What Have We Learned in the
Past Few Years", ANNUAL REVIEW OF POLITICAL SCIENCE 4: 271-93.

*
* 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/



© Copyright 1996–2020 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index