Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


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

st: AR1 correction with xtgls


From   Bryon Balint <[email protected]>
To   "[email protected]" <[email protected]>
Subject   st: AR1 correction with xtgls
Date   Wed, 8 Jan 2014 21:46:11 +0000

Hello, I posted a question about this last week but wanted to restate it more generally. I am using xtgls in Stata 13.0. I am analyzing data for 114 panels, with monthly time intervals. xtgls ran successfully and I got statistically significant coefficients on several variables.

Both a Hausman test and the xtserial command indicated that the data exhibited autocorrelation (rho = 0.6421). I reran the same analysis using the c(ar1) option in Stata. I also had to use the "force" option since the panels are not all evenly spaced. When I did this, about half of the coefficients lost statistical significance. The ones that lost statistical significance were a time trend variable (also spaced monthly), and interactions between other variables and that time trend variable. This leads me to two questions:

1. Is there a reason that the AR1 correction for xtgls would systematically impact some variables more than other variables? (I realize this may be more of a general statistical question than a Stata question)
2. Is it possible that the "force" option is impacting the way that the AR1 correction is applied so that some variables are affected differently than others? I've looked through the Stata documentation but haven't found a full description of how the "force" option is implemented.

Thanks,
Bryon

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index