Re: st: R: Paramertric survival analysis: shared frailty

[Steve Samuels <sjsamuels@gmail.com>]

Victoria,

As far as I can tell, you can't get estimated random effects after a parametric
survival analysis in Stata. Although -streg- will fit shared frailty models, there
is no post-estimation option to create the random effects. As you say, random
effects can be estimated  after -stcox-; the command to do this in Stata 11 and 12
is:

*********************
predict nu, effects
*********************

-stcox-  fits only a random intercept. -glamm- and -xtmepoisson- can
fit more elaborate mixed proportional hazards models and output the random effects.
See Chapter 7 of the the GLLAMM manual. These models are semi-parametric as the are
the Cox models.

I looked through R's contributed packages and the documented examples in HLM and
OpenBugs but nothing to help you there. However SAS's NLMIXED procedure provides a
complete solution if you are willing to write out the likelihood equations.  For
an example see:
http://support.sas.com/documentation/cdl/en/statug/63033/HTML/default/viewer.htm#
statug_nlmixed_sect043.htm.


Steve
sjsamuels@gmail.com



On Jun 23, 2012, at 1:22 AM, <carlo.lazzaro@tiscalinet.it> <carlo.lazzaro@tiscalinet.it> wrote:

Dear Victoria,

shared(varname) is described as shared frailty ID variable in Stata 12.1
manual under -streg- entry (p 359).

Otherwise, see command -streg- Model 2 that leads you to the Shared frailty
ID variable option.

Best regards,
Carlo

-----Messaggio originale-----
Da: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] Per conto di Victoria Allan
Inviato: venerdì 22 giugno 2012 18:49
A: statalist@hsphsun2.harvard.edu
Oggetto: st: Paramertric survival analysis: shared frailty

I conducting a survival analysis and wish to model hospital-level effects.
Using "stcox", there is an option to estimate the random effects
"effect(nu)", however this option is not allowed with the "streg" command
for parametric survival analysis. I wish to know if there is an option
equivalent to "effects(nu)" that can be used with "streg", or alternatively
how else I can get an estimate the random effects?

Kind Regards

Victoria Allan
*
*   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/


*
*   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/


*
*   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/