Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: suggestions for handling *many* ties in duration modelling?


From   "Angeli Kirk" <angeli.kirk@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   st: suggestions for handling *many* ties in duration modelling?
Date   Tue, 12 Sep 2006 12:12:06 -0400

Dear statalist,

I am modelling production uptake times for cash crops (the number of
years from the time snow peas were introduced to a community  as a
potential crop to the year a household chooses to grow them) using a
duration model with time-varying covariates (with the help of the
survival analysis suite). While the span of the observations covers 26
years, more than two-thirds of the 300-household sample take up snow
peas during the first three years, yielding a very large number of
"ties" for uptake time. Another one-sixth do not take up snow peas at
all.

Because we prefer not to make errors in out assumptions about the
baseline hazard function over time, the Cox method would seem to be
the recommended method. However, various websites and manuals mention
"bias" and "bad results" using Cox's exact-partial calculation when
there is a large proportion of ties.

Might anyone explain the "bad results" to me? What are my best
options? Am I better off choosing a parametric distribution?

Many thanks!

-Angeli
*
*   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