on 29/01/2003 3:59 pm, Wagner, Joseph H at [email protected] wrote:
> Thank you for your reply. I'm sorry I didn't make it clear: I am interested
> in the increase in risk across categories of AST
>
>
>
> . stptime if noliverdx==1, by(astlevel) per(1000) dd(4)
>
> astlevel | person-time failures rate [95% Conf. Interval]
> -----------+-----------------------------------------------------------
> AST<0.5 | 1531153.0004 55 0.0359 0.0276 0.0468
> AST 0.5<1 | 1973301.9990 94 0.0476 0.0389 0.0583
> AST 1<2 | 518659.9998 36 0.0694 0.0501 0.0962
> AST 2+ | 137030.0007 28 0.2043 0.1411 0.2959
> -----------+-----------------------------------------------------------
> total | 4160144.9999 213 0.0512 0.0448 0.0586
>
Aha - then I would go for Cox regression. You can use -sts graph- and -sts
list- to plot the risks in the four categories, to confirm that the risk
seems to rise steadily with increasing AST. But I would test the association
between prognosis and AST using the actual data rather than the four
categories. You will then get the hazard ratio associated with a 1-unit
increase in AST, which is useful clinical information.
When using -sts list- you can use the -at- option to limit the output to
survival times of interest rather than seeing them all.
If the relationship is nonlinear, or you suspect it is, you can try
-fracpoly- and see if it identifies a function that is markedly nonlinear.
I like to check fractional polynomials to reassure myself that if the
relationship isn't exactly linear, at least linearity is a good
approximation.
Ronan M Conroy ([email protected])
Lecturer in Biostatistics
Royal College of Surgeons
Dublin 2, Ireland
+353 1 402 2431 (fax 2764)
--------------------
*
* 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/
