Another approach would be to use the time to when the event first occurs for
either eye. Then you would have one observation per subject. We do this sort
of thing in a study of cataract formation as a function of radiation
exposure and other explanatory variables, using maximum opacity betwen the
two eyes.
Al Feiveson
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Ron�n Conroy
Sent: Thursday, November 17, 2005 4:06 AM
To: [email protected]
Subject: Re: st: log rank test for paired data
On 17 Samh 2005, at 05:24, Susan Donath wrote:
> I have a small dataset (n=12 children) with time to event data for
> each eye
> (24 eyes). There are 2 subgroups (of the eyes), group 1 has less
> severe disease, group 2 has more severe disease. A child can have any
> combination of these (eg left eye more severe, right eye less severe,
> etc).
> I can use sts test to test for difference in time to event between
> the 2 groups (using n=24 eyes) but this does not take into account the
> clustered nature of the data.
> Is there any way I can do a log-rank test which takes into account the
> clustering?
You could try -stcox-, which does allow for clustering.
But the problem may run deeper than that. It could be that the status of one
eye may determine the prognosis of the other. An eye with less severe
disease may have a worse prognosis if the child's other eye has severe
disease. (This would certainly be the case in trachoma, for instance, with
the severely diseased eye acting as a nidus for
reinfection.)
And testing a more nuanced hypothesis like this using 24 eyeballs is going
to have your data stretched till they whimper.
Ron�n Conroy
[email protected]
*
* 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/
*
* 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/
