Statalist The Stata Listserver


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

RE: st: Heterokedasticity - help


From   "Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   RE: st: Heterokedasticity - help
Date   Thu, 3 May 2007 19:57:11 +0100

Ingo, Le,

Two thoughts:

First, Stock and Watson have shown that the usual heteroskedastic-robust
SEs for the fixed effects estimator are not consistent, but the
cluster-robust ones are.  See

http://www.nber.org/papers/T0323

Second, -xtivreg2- will run on Stata 8.2 and supports standard fixed
effects estimation with no endogenous regressors as well as -cluster-
and -robust-.

Cheers,
Mark

> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu 
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Tung Le
> Sent: 03 May 2007 19:41
> To: statalist@hsphsun2.harvard.edu
> Subject: Re: st: Heterokedasticity - help
> 
> Hi Ingo, thanks for your help. I use Stata 8.2 Intercool and 
> it does not allows for xtreg,, fe, robust as well as cluster 
> option. For the xtscc, when I run it says: option prop() not 
> allowed (error occurred while loading xtscc.ado) My dataset 
> contains financial information (balance sheets and profit and 
> loss accounts items) of around 730,000 observations of 79,000 
> firms in 12 years. I run xttest3 and it suggests that there 
> is hetero in the sample. Should I trust this tests?
> I also run xttest2 to test for the cross sectional dependence 
> but it says: "firms contains all missing values" - nightmare!
> Le
>  
> 
> ----- Original Message ----
> From: Ingo Brooks <ingo.brooks@gmail.com>
> To: statalist@hsphsun2.harvard.edu
> Sent: Thursday, May 3, 2007 4:09:21 PM
> Subject: Re: st: Heterokedasticity - help
> 
> 
> Le,
> 
> Which Stata version do you use? In Stata 9.2 you can estimate 
> fixed effects models with heteroscedasticity consistent 
> standard error estimates. Just type -xtreg ... , fe robust-. 
> However, typically you should not estimate a fixed effects 
> model with just heteroscedasticity consistent standard errors 
> but rather a model with panel robust standard errors. In 
> Stata 9.2 you can do this by typing -xtreg ... , fe 
> cluster(ID)- where ID is assumed to be the panel identifier.
> 
> However, your panel seems to be a microeconometric panel. 
> Therefore, probably heteroscedasticity is not the main 
> problem for your residuals. Rather the problem might be 
> cross-sectional dependence between the subjects. Therefore, I 
> would recommend you to estimate a fixed effects model with 
> Driscoll-Kraay standard errors since these standard errors 
> are heteroscedasticity consistent and robust to 
> cross-sectional as well as temporal dependence. Thanks to Dan 
> Hoechle's -xtscc- program it is simple to estimate 
> fixed-effects regressions with Driscoll-Kraay standard 
> errors. Just type -xtscc ...
> , fe-. To download the program type -net search xtscc-.
> 
> I hope this helps.
> 
> Best,
> Ingo
> 
> 
> 
> 
> On 5/3/07, Tung Le <empiretung@yahoo.com> wrote:
> > Dear Statalist, I am new to Stata and would like to ask for 
> your help on the heterokedasticity issue:
> > 1. I run xttest3 and find that heterokedasticity exist in 
> my data. However, the author of the test suggests that we 
> should take the test with great care when the data has large 
> N and small T. My data has 730,000 obs over 12 year period 
> (unbalanced). Are they considered small T? and if they are, 
> is xttest3 suitable? it seems that "robvar" test does not 
> allowed for panel.
> > 2. Assume that my data exist the heterokedasticity problem, 
> what type of xt model is suitable to correct this? xtreg..., 
> fe seems not to allow for robust standard error. xtgls? or xtpcse?
> > 3. I tried xtgls but they said matsize is too small. They 
> suggests that matsize should be at least equal to the number 
> of groups of firms in my data so I think they use fixed 
> effects model. However, why does not it require large matsize 
> when I run xtreg..., fe? what is the difference here?
> > Many thanks and hope to receive your advice soon, Le
> >
> > __________________________________________________
> > Do You Yahoo!?
> > Tired of spam?  Yahoo! Mail has the best spam protection around 
> > http://mail.yahoo.com
> > *
> > *   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/
> 
> __________________________________________________
> Do You Yahoo!?
> Tired of spam?  Yahoo! Mail has the best spam protection 
> around http://mail.yahoo.com
> *
> *   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 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index