Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: RE: F statistic after a clustered regression


From   "Mark Schaffer" <[email protected]>
To   [email protected]
Subject   Re: st: RE: F statistic after a clustered regression
Date   Sun, 18 Dec 2005 19:12:47 -0000 (GMT)

Meryle,

I don't think Roger is right in this case, or more precisely, that he has
spotted the source of the difficulty.

Your problem is that the cluster-robust estimator relies on the number of
clusters to go off to infinity for the asymptotics to work.  Moreover, the
cluster-robust variance-covariance matrix won't even be of full rank if
the number of regressors (rows/cols) is greater than the number of
clusters.

If you check your second regression, you'll see you have only 4 clusters. 
Your var-cov matrix is therefore not of full rank, and this is why the
F-test fails.  If you check the regression output from within Stata, there
should be a hyperlink on the failed F-stat that will give you some
details.

The basic point is that you need to have a decent number of clusters for
the cluster-robust var-cov matrix to make any sense, just as you need a
decent number of observations for the coefficient estimator to make any
sense.

Hope this helps.

Yours,
Mark

> Thanks.  This is exactly what I was looking for.
>
> Meryle
>
> ----- Original Message -----
> From: Roger Newson <[email protected]>
> Date: Sunday, December 18, 2005 12:02 pm
> Subject: st: RE: F statistic after a clustered regression
>
>> Stata does not calculate an F-statistic after a clustered
>> regression because
>> the clustered regression uses Huber variances, which are
>> calculated without
>> assuming that the observations are independent and homoskedastic
>> (ie equally
>> variable). The F-test statistic is calculated using the assumption
>> thatobservations are independent and homoskedastic, and is
>> therefore not
>> appropriate for use in a clustered regression analysis. To carry out
>> multi-dimensional hhypothesis tests for a clustered regression
>> model, use
>> the -test- command.
>>
>> I hope this helps.
>>
>> Roger
>>
>> --
>> Roger Newson
>> Lecturer in Medical Statistics
>> Department of Public Health Sciences
>> Division of Asthma, Allergy and Lung Biology
>> King's College London
>>
>> 5th Floor, Capital House
>> 42 Weston Street
>> London SE1 3QD
>> United Kingdom
>>
>> Tel: 020 7848 6648 International +44 20 7848 6648
>> Fax: 020 7848 6620 International +44 20 7848 6620
>>  or 020 7848 6605 International +44 20 7848 6605
>> Email: [email protected]
>> Website: http://phs.kcl.ac.uk/rogernewson/
>>
>> Opinions expressed are those of the author, not the institution.
>>
>>
>> CCarolyn Zhu wrote:
>>
>> Hello listers,
>> Would anybody tell me what it means when stata does not compute an F
>> statistic after a clustered regression? The estimation results
>> with and
>> without the cluster option are provided below.
>> Thanks.
>> . regress loc_fim2_nod katjanine loc_fim1 demented age1 male rand
>> nh_respaidhelp fnnondisp fndisp clin_ab_admt_dum _Isite_2 _Isite_3
>> _Isite_4      Source |       SS       df       MS
>> Number of obs =
>> 464
>> -------------+------------------------------           F( 14,
>> 449) =
>> 18.86
>>       Model |  2495.89099    14  178.277928           Prob > F
>>  =
>> 0.0000
>>    Residual |  4243.41074   449  9.45080342           R-squared
>>  =
>> 0.3703
>> -------------+------------------------------           Adj R-
>> squared =
>> 0.3507
>>       Total |  6739.30172   463  14.5557273           Root MSE
>>  =
>> 3.0742
>> -------------------------------------------------------------------
>> ---------
>> --
>> loc_fim2_nod |      Coef.   Std. Err.      t    P>|t|     [95% Conf.
>> Interval]
>> -------------+-----------------------------------------------------
>> ---------
>> --
>>   katjanine |   .0987764   .0380182     2.60   0.010     .0240608
>> .1734921
>> ? more estimates
>>    _Isite_2 |  -.6510725   .4850485    -1.34   0.180     -1.60432
>> .3021745
>>    _Isite_3 |   -.671119   .5081955    -1.32   0.187    -1.669856
>> .3276179
>>    _Isite_4 |  -.0978212    .585194    -0.17   0.867     -1.24788
>> 1.052238
>>       _cons |   6.814154    1.96921     3.46   0.001     2.944142
>> 10.68417
>> -------------------------------------------------------------------
>> ---------
>> --
>> . regress loc_fim2_nod katjanine loc_fim1 demented age1 male rand
>> nh_respaidhelp fnnondisp fndisp clin_ab_admt_dum _Isite_2 _Isite_3
>> _Isite_4,robust cluster(site)
>> Regression with robust standard errors                 Number of
>> obs =
>> 464
>>                                                       F(  2,
>> 3) =
>> .
>>                                                       Prob > F
>>  =
>> .
>>                                                       R-squared
>>  =
>> 0.3703
>> Number of clusters (site) = 4                          Root MSE
>>  =
>> 3.0742
>> -------------------------------------------------------------------
>> ---------
>> --
>>             |               Robust
>> loc_fim2_nod |      Coef.   Std. Err.      t    P>|t|     [95% Conf.
>> Interval]
>> -------------+-----------------------------------------------------
>> ---------
>> --
>>   katjanine |   .0987764   .0249159     3.96   0.029     .0194828
>> .1780701
>> ? more estimates
>>    _Isite_2 |  -.6510725   .0513326   -12.68
>> 0.001    -.8144356   -.4877094
>>    _Isite_3 |   -.671119   .2526669    -2.66   0.077    -1.475218
>> .1329796
>>    _Isite_4 |  -.0978212   .1077007    -0.91   0.431     -.440573
>> .2449306
>>       _cons |   6.814154    1.29869     5.25   0.013     2.681142
>> 10.94717
>> -------------------------------------------------------------------
>> ---------
>> --
>>
>>
>> .
>>
>> *
>> *   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/
>


Prof. Mark Schaffer
Director, CERT
Department of Economics
School of Management & Languages
Heriot-Watt University, Edinburgh EH14 4AS
tel +44-131-451-3494 / fax +44-131-451-3294
email: [email protected]
web: http://www.sml.hw.ac.uk/ecomes



__________________________________________________________________

DISCLAIMER:

This e-mail message is subject to http://www.hw.ac.uk/disclaim.htm
__________________________________________________________________

*
*   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 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index