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: xtpoisson, fe vce(boot) or xpoisson, fe i(id) vce(boot)


From   "Querze, Alana Renee" <arq@ku.edu>
To   "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
Subject   st: xtpoisson, fe vce(boot) or xpoisson, fe i(id) vce(boot)
Date   Sat, 21 May 2011 19:18:57 +0000

This is my first time posting so I apologize for any list serve faux pas. I am running an analysis on panel data where my dependent variable is a count. I have around 380 districts for 52 months. I have Stata 10.1 running on windows.

My questions is : Does anyone know how to specify a model in Stata that is immune to overdispersion, serial correlation, and heteroskedasticity for panel data where the dependent variable is a count?

Below I have explained everything I've tried so far.

 I tried to check for overdispersion by using estat gof (pois gof for older versions of stata) after running an xtpoisson model. However, I got the error 
 
invalid subcommand gof
r(321);

After searching this list serve I saw that someone else had the same problem. Someone replied that unobserved heterogeneity (which causes overdispersion) is not an issue for longitudinal data. However, I read (in "Fixed Effect Regression Models" by Paul Allison) that time varying omitted variables can still cause overdispersion in a fixed effect model. If anyone can tell me how to test for overdispersion in a xtpoisson model I would appreciate it. 

I decided I could run a model that did and did not control for overdispersion and compare the results. So, I can either run either xtpoisson with a vce(boot) option or xtnbreg. However, in either case I don't know if I'm getting robust standard errors. On this list serve some people said they believed vce(boot) automatically creates robust standard errors. Others said to add i(id) to get robust standard errors, so the code would look like

xtpoisson Dvar Ivar, fe i(id) vce (boot)

I ran my data with and without the i(id) and got very different results (in one my Ivar is significant and in the other it is not).

I also tried  xtpqml but there is no vce(boot) option...

Thank You in advance to anyone willing to tackle this question.

Best,
Alana

*
*   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