RE: st: Simultaneously accounting for clustering at two different levels with vce(cluster) option

["Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk>]

Misha, lilyyor1,

> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu 
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of 
> Misha Spisok
> Sent: Thursday, March 11, 2010 5:39 AM
> To: statalist@hsphsun2.harvard.edu
> Subject: Re: st: Simultaneously accounting for clustering at 
> two different levels with vce(cluster) option
> 
> See, also, Cameron, Gelbach, and Miller, Robust Inference with
> Multi-way Clustering, for which, I believe, the authors have developed
> a Stata package.

I don't think this will work in this applications, for 2 reasons.

First, the Cameron-Gelbach-Miller-Thompson approach is for non-nested clusters (I think that's why they call it "multi-way" rather than "multi-level").  In lilyyor1's application, classrooms are nested in schools.

Second, the cluster-robust approach in general requires the number of clusters to go off to infinity.  8 schools is not very far on the way to infinity!

HTH,
Mark

> 
> On Tue, Mar 9, 2010 at 11:40 AM, Lily Yor <lilyyor1@gmail.com> wrote:
> > Hi, I have data that is clustered at two different levels 
> -- i.e., I have
> > data at classroom level and at school level.  I need to 
> account for these
> > two sets of intraclass clustering, but the clusters at each 
> level are too
> > few to run a formal multi-level model (for example, there 
> are only 8 schools
> > in my data).  Thus, I would like to run a logistic 
> regression model (my
> > dependent variable is binary -- whether a student has 
> passed a certain test
> > or not) and adjust the standard errors by using 
> vce(cluster) option, but
> > there doesn't seem to be a way to simultaneously, in a 
> single model, account
> > for clustering at both levels (i.e., classroom level and 
> school level).
> >
> > The way I have approached this problem thusfar is to run 
> the logistic
> > regression model with vce(cluster) option applied to 
> account for classroom
> > clustering, and then include school dummies in the same 
> model, but this does
> > not seem to be a satisfactory solution:
> >
> >  logit [dep var] [ind_var1] [ind_var2] [ind_var3] [school2] 
> [school3]
> > [school4] [school5] [school6] [school7] [school8], 
> vce(cluster classroom_id)
> >
> > If you have any tips and suggestions on how I could use the 
> vce(cluster)
> > option to account for both classroom and school clustering, 
> I would very
> > much appreciate your help.
> >
> > Thank you so much.
> >
> > *
> > *   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/
> 
> *
> *   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/
> 


-- 
Heriot-Watt University is a Scottish charity
registered under charity number SC000278.


*
*   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/