# RE: st: Variance decomposition

 From "Nick Cox" To Subject RE: st: Variance decomposition Date Sun, 11 Mar 2007 18:02:54 -0000

```Stata itself can help you in these queries.

. search variance components

yields, among other pointers,

[XT]    xtmixed . . . . . . . . . . Multilevel mixed-effects linear regression
(help xtmixed)

SJ-6-1  st0095  . . . . . . . . . . .  Estimating variance components in Stata
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . Y. Marchenko
Q1/06   SJ 6(1):1--21                                    (no commands)
describes using xtmixed to estimate variance components
in linear models

SJ-4-4  st0077  . . CIs for the variance comp. of random-effects linear models
(help xtvc if installed)  . . . . . . . . . .  M. Bottai and N. Orsini
Q4/04   SJ 4(4):429--435
confidence intervals for the variance components of
random-effects linear regression models.

Nick
n.j.cox@durham.ac.uk

Michael Crain

> This methodolgy is different from ANOVA.
>
> Under a variance components approach, the variance of each of
> the factors is
> independently generated by a random process.  Once the
> variance of a factor
> has been estimated, it remains fixed for the remainder of the
> analysis.  The
> variance components approach uses zero means and constant but unknown
> variances for all the factors.  (Only the error term meets
> this criteria
> under ANOVA.)  Therefore, each factor can be disentangled
> from the other
> factors to explain each one independent of the others.  The
> substance of
> this approach is to explain that the differences among the
> factor variances
> are natural and independent of the other effects.  The
> ultimate result of
> this analysis is expressed as a measure of relative
> importance of each
> factor.  ANOVA allows us to test statistical significance but
> does not allow
> one to find relative importance.

*
*   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/
```