st: variance partitioning in a 3-level model (fwd)

["Cathy L. Antonakos" <cathya@umich.edu>]

I asked a question (included below) about partitioning variance in a 3-level model using -xtmixed- and have since found the answer. It is the option "variance" added at the end of the line, as follows (in a fully unconditional model). Very simple.

.xtmixed t_rc || hosp_id: || unit_id:, variance mle

I found the article by Yulia Marchenko, Stata Journal vol. 6 no.1, very helpful.
Cathy

---------- Forwarded message ----------
Date: Wed, 3 Jan 2007 00:41:17 -0500 (EST)
From: Cathy L. Antonakos <cathya@umich.edu>
To: statalist@hsphsun2.harvard.edu
Cc: Cathy L. Antonakos <cathya@umich.edu>
Subject: variance partitioning in a 3-level model

A few weeks ago, I asked re: regression with clustered data, and was steered 
toward a mixed model. I am working with -xtmixed- now. There are hospital level 
(level 3) and unit level (level 2) effects on people (level 1), and both unit 
and hospital are easily accommodated with -xtmixed-.

The -xtmixed- command I am using is something like this:

.xtmixed t_rc no_pes t_cweq ||hosp_id: ||unit_id:, mle

In a two-level model, I would use -loneway- to get the intra-class correlation 
for level-2 effects (fully unconditional model).
I did that here,

.loneway t_rc unit_id

I also did that for level-3 as follows, but am not sure this is correct:

.loneway t_rc hosp_id

In Bryk & Raudenbush (1992), in the chapter on 3-level models, the "proportion 
of variance among" level-2 units (classrooms) is described as "variance among 
classrooms within schools." This language is confusing to me, as I am guessing 
that the proportion of level-2 variance in a fully unconditional model does not 
account for level-3, except in the denominator. Is that correct? Or should I be 
running the fully unconditional 3-level model to get these variance estimates?

Thanks in advance.

Cathy



*
*   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/