Re: st: mixed

[Clyde B Schechter <clyde.schechter@einstein.yu.edu>]

Erik Carter wrote about his quest to calculate an intra-class correlation using -xtmixed- and -quickicc-.

The latter is a user-written command, not part of official Stata, and I have non familiarity with it.

But from the description of his problem, the simplest route to his answer would be:

xtset study_id
xtreg qt

The output will include the ICC as rho.

Another route to the ICC is via the -loneway- command.

There is the other question of why -xtmixed- is not converging.  This may reflect a problem in his data that is hard to comment about without seeing the output.   Assuming that his iteration list is getting stuck somewhere with the log likelihood not changing over further iterations, he can rerun the -xtmixed- command specifying a fixed number of iterations to see how the values are looking at the point where progress ceases to be made.  Another way to do this is to add the -trace- option.  Typically in this situation one of the variance components is marching to zero (which shows up in the trace output as one of the _ln_sig_*'s marching to negative infinity.)

Hope this helps.

Clyde Schechter
Department of Family & Social Medicine
Albert Einstein College of Medicine
Bronx, NY, USA

Bronx, NY, USA


*
*   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/