Statalist The Stata Listserver


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

Re: st: Estimating a model allowing for AR(1) in residuals with weights in panel


From   David Jacobs <[email protected]>
To   [email protected]
Subject   Re: st: Estimating a model allowing for AR(1) in residuals with weights in panel
Date   Wed, 12 Apr 2006 14:16:38 -0400

Check out the cites to the articles by Beck and Katz in the manual entry on xtpcse. They claim that xtgls shouldn't be used because its estimates of the standard errors are far to optimistic. And I've certainly found comparatively enormous t-values when I've used xtgls, so I suspect B & K are right.

By the way, at least xtpcse and xtgls assume that you have at least ten to twelve time periods. I don't remember if there are similar strictures on xtregar.

Dave Jacobs

At 02:00 PM 4/12/2006, you wrote:

Dave,

Thanks!

I have also seen xtgls. Indeed what are the differences between xtpcse? Conceptually are there any difference in the adjustment method than xtregar?

Thanks!

Tak Wai Chau

David Jacobs wrote:

Xtregar fe does allow for aweights (and probably fweights) if you use two of the options for estimating the AR1 term.

Check the help file or the manual.

Probably, however, you will find that xtpcse will better suit your needs if you have a large enough T for this estimator. Note that Beck and Katz do NOT recommend the PSAR1 option in xtpcse that estimates different ar1 corrections for each case (or state in your study).

Dave Jacobs

At 09:50 AM 4/12/2006, you wrote:

Hi, Statalist users,

I have a question about estimating a model allowing for AR(1) in residuals with weights.

I have a dataset with state-year level data. The model is like this:

y_it= a + b*policy_it + c_i + d_t + u_it

where i stands for states and t states for year. policy is a policy implemented at different time in different states. c_i are state dummies (all states except one), and d_t are year dummies (all year except one), thus it is a difference in difference model. I also want to do this regression with state population size as weights.

If u_it is serially correlated for each state, and I would like to allow for AR(1) for this u_it over time for each state to obtain parameter estimates, what should I do in Stata?

I have thought of xtregar, fe, but it does not allow weights.

BTW, I think the convention is that we have the autoregressive parameter the same across all states. I wonder if it is identified if I allow different autoregressive parameters in different states.

Thank you very much in advance!

Tak Wai Chau

*
* 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/



*
*   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 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index