st: time-varying treatment for survival analysis

[László Sándor <sandorl@gmail.com>]

I have a project where people accumulate treatment (varying dosage)
over time. I am a bit confused whether it is OK to -stsplit- the
sample at time of snapshots of accumulated treatments, merge in the
snapshot-specific records, and do "standard" [ST] analysis on that.

I am OK assuming that treatment accumulated by the time of a snapshot
is what drives the hazard afterwards, at least until the next
snapshot.

Is it OK to cut the data by dosage (using the -by- option) for
survival and hazard graphs, e.g.?

If an example helps, think of the daily mortality hazard over
1985-2010 where I observed annual snapshots of wealth lost or won on
the stock market over a year for 8 years (2000-2007) which I can also
accumulate over time, of course — plus plenty of covariates at even
more annual snapshots.

(By the way, I observe day of death, I only observe year of birth, I
hope an approximation assuming all being born on July 1st is
defensible.)

Thanks!

*
*   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/