st: stsetting overlapping multiple spells

[László Sándor <sandorl@gmail.com>]

Hi all,

I use Stata 11.1 on Unix and the Mac, and I have a question on
specifying my data structure for some survival analysis.

I have data on tasks for thousands of individuals, invited and due at
various time over a year. I am interested in task completion for a
specific kind of job, where any other job pending might distract the
individual (crowd out effort). As far as I understand, this is not a
typical competing risk scenario, as failure can occur (and I observe
it) for all spells separately. Most tasks are completed in my case.
Earlier completion of one task does not obstruct any other.

I hoped to allow for effects of a task due soon (say, in 5 days)
decreasing the hazard of finishing another task of the same person. My
fear is that this requires heroic stsplitting to be able to link the
right calendar times across various spells (having different event
times otherwise) --- I might need to drop a lot of fine variation in
job completion time to realistically be able to do that.

Yet even having done some stsplitting (by days, say), I am not sure
how to proceed. Is there any better way than some crude stsplitting in
calendar time and allowing some "peer effects" (as if tasks were
individuals and the worker corresponded to "classes")?

Any help would be greatly appreciated.

Some mock-up example data could look like the following:

PERSON     INVITED     DUE             FINISHED
1                 3/1 8:12      4/10 0:00       4/18   23:46
1                 3/15 11:52  4/24 0:00       4/10   21:13
2                 3/2 2:35      4/11 0:00       3/3    12:14
...

Thank you very much!

Laszlo

László Sándor
PhD candidate in Economics
Harvard University

*
*   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/