Hi,
Sorry for my last question, please forget it.
Kit saids
> There is no obvious way of considering a test for autocorrelation in
> the pooled OLS context. You would have to conduct a test on each
> panel, and determine some way of combining results from those tests
> (similar to the issue of designing a panel unit root test). That is
> why I suggested using ivreg2 (in the absence of unobserved
> heterogeneitty) and estimating the model with HAC standard errors. If
> HAC standard errors are quite similar to pooled OLS standard errors,
> then there is no concern of non-iid disturbances.
I got an error message after I use ivreg2;
Error: estimated covariance matrix of moment conditions not of full rank;
overidentification statistic not reported, and standard errors and
model tests should be interpreted with caution.
Possible causes:
covariance matrix of moment conditions not positive definite
covariance matrix uses too many lags
singleton dummy variable (dummy with one 1 and N-1 0s or vice versa)
fwl option may address problem.
What's the meaning of this error message? Is there another way to test
for autocorrelation?
Regards,
A. Sura
On 4/21/07, Keynes M. Smith <[email protected]> wrote:
Hi,
I've problem with the command -tsset-.
When I try to set time variable, it shown that
. tsset year, yearly
repeated time values in sample
So I can not use some command such as -levinlin-.
I don't understand why I can't set only time variable because when I
use -tsset- to set both
time variable and panel id variable, it work.
tsset country Year, yearly
panel variable: country, 1 to 10
time variable: Year, 1990 to 2005
What's wrong with my data?
Regards,
A. Sura
On 4/21/07, Kit Baum <[email protected]> wrote:
> Keynes said
>
> I already run the fixed effect model and F-test accept the null
> hypothesis.
>
> So I have to work on pooled OLS.
>
> I also check for the heteroskedastic, here is the result
>
> Breusch-Pagan / Cook-Weisberg test for heteroskedasticity
> Ho: Constant variance
> Variables: fitted values of gdppercapita
>
> chi2(1) = 2.58
> Prob > chi2 = 0.1085
>
> which, if I'm not misunderstand, mean that there's no heteroskedastic
> problem.
>
> For the P-value, is that mean I have to divide by 2 before check the
> significance of the
>
> coeffiicent?
>
> Could you please suggest me how to test autocorrelation and how to
> correct it if I found the
>
> problem.
>
>
>
>
> No, you do NOT have to divide the pvalue by two. That would only be
> appropriate if you were conducting a one-tailed test. For a two-
> tailed test, leave it alone.
>
> There is no obvious way of considering a test for autocorrelation in
> the pooled OLS context. You would have to conduct a test on each
> panel, and determine some way of combining results from those tests
> (similar to the issue of designing a panel unit root test). That is
> why I suggested using ivreg2 (in the absence of unobserved
> heterogeneitty) and estimating the model with HAC standard errors. If
> HAC standard errors are quite similar to pooled OLS standard errors,
> then there is no concern of non-iid disturbances.
>
>
> Kit Baum, Boston College Economics
> http://ideas.repec.org/e/pba1.html
> An Introduction to Modern Econometrics Using Stata:
> http://www.stata-press.com/books/imeus.html
>
>
> *
> * 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/
