st: Detecting unequal variance in rates of change - xtintreg, xtmixed, or cmp?

["Seed, Paul" <paul.seed@kcl.ac.uk>]

Dear Statalist ,

I have a slightly complex hypothesis to test, 
and would appreciate some advice about how to go about it.

I have repeated measurements at various times in pregnancy (no consistent pattern)
of a blood marker that may be related to early delivery.
Specifically, we think that it might rise over time in some women who are at higher risk, 
as part of a protective mechanism; but not in others for whom the protective mechanism 
may not work so well.  As a further complication, the marker is left-censored, 
and log-transformed for Normality. 

The basic model is:
               iis patient_id
               xtintreg  l_xxx_lb l_xxx_ub i.case##c.gest_sample  l_pool_xxx , nolog

Here xxx is the marker; l_xxx_lb l_xxx_ub are the logged values, with _ub and _lb showing the upper and lower limits of censoring, 
and gest_sample is the gestation at which the blood sample is collected.

This allows me to test whether the average slopes are different between cases and controls; but not whether the 
slopes are more varied in the cases than in the controls.

I think some use of -xtmixed- [Standard Stata command, which does not accept censored data]  or -cmp- 
[Copyright David Roodman 2007-13, which does] may give me the answer; but I may have to create 
appropriate dummy variables first, instead of relying on Stata to automatically create them.

Any thoughts gratefully received. 


Paul T Seed, Senior Lecturer in Medical Statistics, 
Division of Women's Health, King's College London
Women's Health Academic Centre, King's Health Partners 
(+44) (0) 20 7188 3642.




*
*   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/