Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# st: From: Marc Peters <marcpeters1002@gmail.com>

 From owner-statalist@hsphsun2.harvard.edu To statalist@hsphsun2.harvard.edu Subject st: From: Marc Peters Date Fri, 13 Dec 2013 18:01:35 +0000

```Dear Statalist users,

I am new to multilevel modeling and I am currently trying to estimate
a model using xtmixed.

Most of my models look reasonable, but when I add a specific (highly
significant) level-2 variable (var2) the level-2 variance component
becomes strangely small.

I use the following command where var1 is a level-1 variable and var2
and var3 level-2 variables.

xtmixed depvar var1 var2 var3||id:, var ml cov(un)

Computing standard errors:

Mixed-effects ML regression                     Number of obs      =        78
Group variable: id                              Number of groups   =        28

Obs per group: min =         1
avg =       2.8
max =         5

Wald chi2(3)       =     89.55
Log likelihood =  35.123461                     Prob > chi2        =    0.0000

------------------------------------------------------------------------------
depvar |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
var1 |   .2780454   .0370738     7.50   0.000      .205382    .3507087
var2 |   .0038315   .0009236     4.15   0.000     .0020213    .0056416
var3 |  -.3500793   .0929506    -3.77   0.000    -.5322591   -.1678995
_cons |   .6950143   .0682743    10.18   0.000     .5611993    .8288294
------------------------------------------------------------------------------

------------------------------------------------------------------------------
Random-effects Parameters  |   Estimate   Std. Err.     [95% Conf. Interval]
-----------------------------+------------------------------------------------
id: Identity                 |
var(_cons) |   6.95e-21   6.63e-20      5.19e-29    9.31e-13
-----------------------------+------------------------------------------------
var(Residual) |   .0237903   .0038095      .0173819    .0325612
------------------------------------------------------------------------------
LR test vs. linear regression: chibar2(01) =     0.00 Prob >= chibar2 = 1.0000

If I have understood it correctly, the variance component for id
should not become smaller when adding a significant level-2 variable
to the equation. Or am I completely mistaken?

Thank you,
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/
```