The Stata manual suggests that the intraclass correlation coef (ICC) in a logistic hierarchical mixed model ('multilevel' mixed model) can be computed as
sigma_panel/(sigma_panel + 1)
after running the xtlogit command with appropriate data. For instance see p 378 ref volume Su-Z for release 6.
I am new to Stata, and from all other work, I would expect pi^2/3 rather than the 1 in the denominator for the ICC. This because the error variance on the logistic scale is pi^2/3. I would accept the 1 if the link function was probit rather than logit.
Does Stata use a constraint I am unaware of contained in the quadrature calucaltion? It appears it might from formula on page 381 of manual referred to above. Or am I just missing something?
Thanks to anyone who knows.
Michael DeLorenzo
Maine Medical Center
*
* 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/
