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: Cluster size
From
Stas Kolenikov <[email protected]>
To
[email protected]
Subject
Re: st: Cluster size
Date
Fri, 11 Mar 2011 14:42:32 -0600
On Fri, Mar 11, 2011 at 12:59 PM, C Butler <[email protected]> wrote:
> I am estimating a multiple logistic regression model and would like to
> adjust the ORs by client clustering at particular facilities using
> Stata's "cluster" option. I have about 75 clusters; about 8 of them
> have 1-10 observations, and the vast majority have over 200. What,
> ideally, is the minimum cluster size for estimates to be consistent?
The size of the cluster usually may have bearing on biases of variance
components in multilevel models, but as long as you estimate a simpler
model, I don't think the number of observations in a cluster should
matter that much. A size measure that matters a lot for the standard
errors and normal confidence intervals to work reasonably well is the
number of clusters, and 75 looks reasonably large (unless you have 50
regressors in the model). Finally, consistency is an asymptotic
property, so what you asking is essentially, how many angels fit on
the tip of a needle. A more reasonable question to ask is, what is the
size of the data set, in terms of all of # of clusters, # of obs /
cluster, and degree of imbalance in cluster sizes, for estimates to
have acceptable biases and acceptable closeness of the sampling
distribution to the normal one. My intuition in your case would be
that you are probably OK, but other folks with better knowledge may
jump in if there is specific simulation evidence that can be
mentioned.
--
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: I use this email account for mailing lists only.
*
* 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/