Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


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

st: RE: RE: RE: RE: Autocorrelation test for pooled cross section data


From   "Elizabeth Dhuey" <elizabeth.dhuey@utoronto.ca>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: RE: RE: RE: Autocorrelation test for pooled cross section data
Date   Wed, 29 Sep 2010 11:37:54 -0400

Due to the length of my time period, I have some concern regarding
correlated errors over time. I have plotted the residuals and see some
evidence of correlation over time. 

However, my attempt to find a formal test that produces a p-value is based
on a request from a journal referee. My main issue is that I can't seem to
find any formal test that is appropriate for pooled cross sectional data. My
current thought is to calculate a durbin-watson statistic for each state and
report the distribution of the statistic. However, I suspect that someone
has already formalized a produce for this kind of data structure.  

Elizabeth


-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Nick Cox
Sent: September-29-10 6:09 AM
To: 'statalist@hsphsun2.harvard.edu'
Subject: st: RE: RE: RE: Autocorrelation test for pooled cross section data

A minimal but presumably useful check is to calculate some flavour of
residuals and plot them against cohort identifier and also space. 

As a non-economist -- and in one very strained sense an ex-economist -- I
don't have to sign up to the notion that every check should be matched by a
formal test producing a P-value.  

I don't know off-hand how worried you should be relatively about correlation
in time and in space for your kind of data, but in principle both could be a
concern. 

Nick 
n.j.cox@durham.ac.uk 

Elizabeth Dhuey

I apologize if I don't understand your questions but here is my best shot at
answering them. 

I'm trying to test whether my errors are correlated across time (i.e. across
cohorts). I have a problem because I have 50 different states so I can't use
a simple Durbin-Watson statistic because I could only use that if I had only
one state. 

I only include dummy variables for the states and do not take into account
the location or any spatial data regarding the states. I only allow for an
intercept change for each state. 

Nick Cox

I am unclear on what you want, but as I understand it Durbin-Watson tests
apply only to time series. 

If you have data on 50 U.S. states and are treating them as spatial series,
which of your variables encode information of the location, contiguity,
whatever, of those states? If it is -birthstate- as a categorical variable,
how is that spatial? 

Elizabeth Dhuey

I'm trying to figure out what is the most appropriate test to use is when
working with pooled cross sectional data. In particular, I have aggregated
individual level measures to state level averages for different cohorts from
the U.S. Census. 

I can't use estat dwatson because I have multiple panels and I don't believe
that I can use xtserial because I don't have a true panel.  

I am running the following regression:

xi: reg Y X i.birthstate i.cohort [pw=wgt], robust cluster(birthstate)


*
*   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/

*
*   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   |   Site index