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 at the end of May, and its replacement, statalist.org is already up and running.


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

Re: st: Autocorrelation and heteroscedasticity following xtreg


From   Nick Cox <njcoxstata@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Autocorrelation and heteroscedasticity following xtreg
Date   Wed, 19 Oct 2011 12:58:07 +0100

As you are now discovering some fundamental properties of the data
that might have implications for your modelling.

I doubt that there is a easy and quick fix to your problem. To get the
tests you want would seem possible only by fudging your data, which is
clearly not advised.

Conversely, you may get something useful out of looking at the
correlation between something and L.something, and similarly for other
lags. Similarly although there is a bias in some disciplines that the
only way to get a handle on something is to apply a pre-canned
significance test, heteroscedasticity can often be investigated by
plotting variances or equivalently (something)^2 and thinking about
how far it is or is not constant in practice. This can be as or more
illuminating than some portmanteau-omnibus-factotum test.

Naturally I have no idea precisely what "something" is or should be in
your case; this is just a very general prescription.

Nick

On Wed, Oct 19, 2011 at 10:50 AM, lreine ycenna <lreine.ycenna@gmail.com> wrote:
> In fact, I cannot run -xtregar - as it also tells me that there is no
> observation. I use -tsreport, panel  list - to check the gap, and
> apparently I have 1000 gaps....would I have to take out all these gaps
> in order to run these tests?
>
>  tsset cn year
>       panel variable:  cn (unbalanced)
>        time variable:  year, 1977 to 2002, but with gaps
>                delta:  1 unit
>
> . tsreport, report panel
>
> Number of gaps in sample:  1100
>
>
> On 19 October 2011 10:39, Scott Merryman <scott.merryman@gmail.com> wrote:
>> Yes.
>>
>> -xtserial- will run -reg d.(var1 var2..)- but if there are enough gaps
>> then this will fail to run.  Is the -delta()- option from -tsset-
>> correct in your case?
>>
>> Scott
>>
>>
>> On Wed, Oct 19, 2011 at 4:30 AM, lreine ycenna <lreine.ycenna@gmail.com> wrote:
>>> yes with gaps it says. do you think it's the problem?
>>>
>>> On 19 October 2011 10:27, Scott Merryman <scott.merryman@gmail.com> wrote:
>>>> Does -tsset- report gaps in the time variable?
>>>>
>>>> -xtregar, lbi- reports a modified Durbin-Watson statistic.
>>>>
>>>>
>>>> On Wed, Oct 19, 2011 at 4:02 AM, lreine ycenna <lreine.ycenna@gmail.com> wrote:
>>>>> After setting my panels using tsset, I typed xtserial x y z - the 'no
>>>>> observation' r(2000); pops up. I searched the statalist forum and It
>>>>> seems that many people have experienced this issue, but there's yet a
>>>>> solution to this. It has nothing to do with the data being unbalanced
>>>>> as I've tried xtserial both on an unbalanced panel and a balanced one.
>>>>> All my data are numerical, so there's no problem with that. Any idea?
>>>>>

*
*   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