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]

st: Additional Intraclass correlation for 2nd level clustering


From   Laurie Molina <molinalaurie@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   st: Additional Intraclass correlation for 2nd level clustering
Date   Wed, 2 May 2012 12:48:59 -0500

Hi,

I have a 4 million observations data set, but I suspect there is
intraclass correlation among some groups.
To clarify, my data does not come from a survey, but from a regulatory
report. So that I have data from the clients of cell phone companies.
I can identify the firm providing the service, as well as the tarifary
plan assigned to each observation. This is not a sample, but the
population.
On a previous thread  some listers suggested me to use clustered
standard errors. I think the most appropiate would be to cluster at
the highest level (.50 intraclass correlation reported using loneway),
however, If i do that i will get only six clusters with a highly
unbalanced number of observations in each one, so the consistency of
my estimates (that if i understand correctly depends on the number of
clusters and not in the number of observation) may not be achieved.
Something similar may happen if I do a multilevel estimation, as again
the higuest level number of groups is very small (6).
So i would like to cluster by tarifary plan (.41 intraclass
correlation reported using lonewat) because i would get more than one
hundred groups. However, as each tarifary plans belong to one and only
one company, there will still be some correlation among obsevations
from tarifary plans of the same firm.

So my questions are mainly two:

1. Is there any way to measure the intraclass correlation that i will
not be capturing when clusteting between tarifary plans and not firms?
Something like the intraclass correlation asociated with clustering by
firms, additional to the intraclass correlation already captured when
clustering by tarifary plans.
2. How can I know if the INTER class correlation among observations
from the same firm, remaining after clustering by tarifary plan, is
high enough so that i should be concerned?

Thank you all in advance, and best regards!
LM
*
*   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 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index