Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


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

Re: st: How to test to heteroskedasticity after -vce(cluster clusvar)- command?


From   Ricky Lim <[email protected]>
To   [email protected]
Subject   Re: st: How to test to heteroskedasticity after -vce(cluster clusvar)- command?
Date   Sun, 25 Aug 2013 03:25:31 +0800

Hi Federico,

Thank you very much for your reply & the references.

May I know how do we normally decide whether we should use two way or
multiple way clustering?
I understand that one way clustering is necessary when data is
collected from 'clusters' e.g. different household, different states,
etc.
But within the data there are many 'clusters' as well,
such as different occupation groups, different years of birth,
different years of data, etc.

I am actually analysing individual level data for a number of
organisations across 5 years.
Each year different samples of individuals are used and is an
independent cross sectional data.
I'm using the earliest year as a baseline to compare with subsequent
years to see whether there is a difference in the dependent variable..

Based on what I have read so far,
I think for my data it is fine to use one way clustering.
I have year dummies in my model, which I think should absorb the common shock.

Would you or any Statalisters be able to advise further?

Thank you very much in advanced.

Regards,
Ricky


On 25 August 2013 00:15, Federico Belotti <[email protected]> wrote:
> When the regression errors are correlated within groups the assumption of independent errors will be violated and the unadjusted OLS standard errors often have a substantial downward bias. -vce(cluster clusvar)- usually solves this issue allowing a valid inference. See for example
>
> Moulton, Brent R., 1986. "Random group effects and the precision of regression estimates," Journal of Econometrics, vol. 32(3), pages 385-397.
>
> So, once you have clustered your standard errors in the right way (Do you have a multiple-clustering structure in your data? see Baum, Schaffer & Stillman's -ivreg2- or Cameron, Gelbach & Miller -cgmreg-) you shouldn't need to test for heteroskedasticity in a linear regression framework.
>
> Hope this helps
>
> Federico
>
> Il giorno 24/ago/2013, alle ore 05:38, Ricky Lim <[email protected]> ha scritto:
>
>> Dear Statalisters,
>>
>> Hope you can provide some advices.
>>
>> I ran a multiple linear regression using -vce(cluster clusvar)-
>> command for my cross sectional survey data.
>>
>> When I tried to use -hettest-,
>> this appears:
>> "hettest not appropriate after robust cluster()"
>>
>> Does this mean that it is not necessary for me to test for
>> heteroskedasticity after clustering my standard errors?
>> Or is there another way to do it?
>>
>> Could not find a similar question in the archive hence I posted it here.
>>
>> Thank you very much in advanced.
>>
>> Regards,
>> Ricky
>> *
>> *   For searches and help try:
>> *   http://www.stata.com/help.cgi?search
>> *   http://www.stata.com/support/faqs/resources/statalist-faq/
>> *   http://www.ats.ucla.edu/stat/stata/
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index