Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: reliability with -icc- and -estat icc-

From   "JVerkuilen (Gmail)" <>
Subject   Re: st: reliability with -icc- and -estat icc-
Date   Tue, 26 Feb 2013 23:41:13 -0500

On Tue, Feb 26, 2013 at 9:26 PM, Lenny Lesser <> wrote:
> so I did the model the way Rebecca suggested and get an ICC of 0.34
> (with Rator as a class)

> Then I did it the way you did it and got an ICC as 0.22 (with Rator
> not class, but with covariance as independent)

> I'm concerned in your model the Rator is not a class/dummy variable.

I fit the analog of the randomized block design, with random Rator but
fixed Applications.

. xtmixed Score i.Application if Rator != 4, || Rator:, covariance(identity) var

If you fit it as a class variable but omitting Rator 4 (which I just
tried) it pushes the R.Rator random effect to 0. I took Rebecca's code
(minus the little typo of Rator spelled as Rater) and simply added the
relevant if:

. xtmixed Score i.Application if Rator != 4, || _all: R.Rator,

(Notice that both models exclude Rator 4.)

These two models are statistically equivalent because they have the
same residual log likelihood. I'm not sure why the _all: R.Rator is
used here. Rebecca, could you chime in? I'm not saying you're wrong.
Sometimes there are different ways of computing the same model. (I
need to reread the manual.)

JVVerkuilen, PhD

"It is like a finger pointing away to the moon. Do not concentrate on
the finger or you will miss all that heavenly glory." --Bruce Lee,
Enter the Dragon (1973)
*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index