Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: problem with shared frailty cox model


From   rgutierrez@stata.com (Roberto G. Gutierrez, StataCorp)
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: problem with shared frailty cox model
Date   Mon, 17 May 2004 13:03:53 -0500

Marcus Fischer <mafische@mcw.edu> asks:

> I tried the shared frailty option in a cox proportional hazard model to
> account for the within-family correlation of my pedigree data.  After
> several hours (!!!) I got a convergence error when estimating the frailty
> variance.  This just happened in a larger dataset (for example in 1500
> individuals from 600 families, 1-4 members within a family); in a smaller
> subset it's working fine. Is it a memory problem? Are there any
> requirements for the coding of the shared frailty variable ?

With 600 families, there are no memory, coding, or matrix-size limitations
issues, although the calculation can be slow.  In its current implementation
the shared frailty model is fit via penalized likelihood, which involves 
the repeated fitting of standard Cox models with dummy variables for the 
families, hence the matrix calculations can get quite large.

As to why the estimation does not converge, there could be many reasons for 
this, as with all models.  One possibility is that you have near collinearity
in the regressors.  If this is the case, then standard Cox would converge
just fine, but iterative Cox model fits with differing hypothesized frailty 
variances could fail to converge sometimes.  As such, the overall model would
not converge.

If Marcus wishes, he can email me his data and model specification, and I 
may be able to better diagnose the problem.

--Bobby
rgutierrez@stata.com
*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index