In a regression analysis, unless the predictors are uncorrelated, the percentage of the variance attributed to a given independent variable cannot be ascertained uniquely. In the circumstance you describe, patients are presumably not distributed randomly among practices and trusts. Consequently the decomposition you seek is not available. David Greenberg, Sociology Department, NYU
----- Original Message -----
From: sdm1 <[email protected]>
Date: Friday, April 3, 2009 3:26 pm
Subject: st: Decomposing cost variations between individual and higher level effects
To: [email protected]
> I have the annual hospital costs for 5 million patients. Each patient
> can
> be allocated to one of 36 age/gender groups. All patients are
> attached to a
> general practice (n=8,000), and all practices are located within an
> administrative area, a primary care trust (n=152). In other words, patients
> are nested within practices which are nested within trusts.
>
> How can I decompose the total variation in costs between those attributable
> to the individual's age/gender, those attributable to the practice, and
> those attributable to the trust?
>
> Thanks.
>
> Steve
>
> *
> * 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/
