|
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: Re: R: Survival Analysis Censoring stset
Dirk--
For how many years do you have outcomes? "Interval-censored" data
methods are, in my opinion, best applied to data in which there are
many possible time points. I think your data are more properly
classified as "grouped" or "discrete". With a relative small number
of years of observation, I suggest calendar time as a time-scale.
This topic has come up before ( http://www.stata.com/statalist/
archive/2006-11/msg00473.html ).
I haven't reread the thread, but I suggest that you try a survival
model appropriate for discrete data, either -hshaz- or -pgmhaz8- by
Stephen Jenkins (-findit hshaz-/ -findit pgmhaz8-). Control for
company age with covariates.
Note that "left-censored" means the outcome for a company took place
before the start of observation. You have "left-truncated" or
"delayed entry" data: some of your companies are "survivors" of
original cohorts of companies which started before the study.
-Steve
On Sep 18, 2008, at 11:57 AM, Dirk Volland wrote:
Hello again,
thank you Carlo. It seems that my description has been a little
confusing. So here another try:
(1) The structure of the data with multiple episodes for each id is
id | year0 | year1 | failure
1 | 2002 | 2003 | 0
1 | 2003 | 2004 | 0
1 | 2004 | 2005 | 1
stset with the command
stset year1, id(id) failure(failure)
With no gaps I meant that annual data is always available (2002 2003
2004 ... without missing years). Is this data interval censored just
because it is discrete and the failure event can only be measured at
year1 meaning the exact time between year0 and year1 is not available?
Is this handled by Stata in an appropriate manner?
(2) When I stset data with the command
stset year1, id(id) origin(time founding) enter(time start)
Here founding contains the time when the subject first came at risk
and start contains the time the observation window starts. So in a way
I know when a company was founded, but I want to start analysis in a
certain year. Are these observations still left censored (in
comparison to just deleting records that are prior to the observation
window)? Is this handled by Stata?
Thanks a lot,
dirk
2008/9/18 Carlo Lazzaro <[email protected]>:
Dear Dirk,
as far as your first question is concerned:
- it seems to me that your following statements "time span as 2006
and 2007
without gaps" and "the exact time between year0 and year1" conflate.
Should your analysis start in 2006 (that is, t0)and stop in 2007
(that is,
t1), some companies will have gone to bankruptcy (that is
failure=1) and
other will have survived (that is, failure=0). The latter are
right-censored
observations, since they might go to bankruptcy (or not) in the
future and,
anyhow you will not know, since you have stopped your analysis.
- stset - will automatically account for censored observations.
For further details on this topic, I will recommend you to take a
thorough
look at:
Cleves MA, Gould WG, Gutierrez R. An Introduction To Survival
Analysis
Using Stata. 2nd edition. College Station: StataPress, 2006;
[ST] Stata manual. Survival analysis and epidemiological table.
Release 9
Two other relevant contributors of the Statalist - Maarten Buis
(http://home.fsw.vu.nl/m.buis/) and Stephen Jenkins
(http://www.iser.essex.ac.uk/teaching/degree/stephenj/ec968/
index.php.)
published really interesting papers as well as teaching-notes on
the topics
you are interested in.
HTH and Kind Regards,
Carlo
-----Messaggio originale-----
Da: [email protected]
[mailto:[email protected]] Per conto di Dirk
Volland
Inviato: gioved� 18 settembre 2008 15.29
A: [email protected]
Oggetto: st: Survival Analysis Censoring stset
Hello,
i have to quesitons regarding censoring and how it is handled in
Stata's survival package.
(1) When I have the following data structure:
id | year0 | year1 | failure
stset with the command
stset year1, id(id) failure(failure)
and year0 and year1 always represents a time span as 2006 and 2007
without gaps. Is this data interval censored just because it is
discrete and the failure event can only be measured at year1 meaning
the exact time between year0 and year1 is not available? Is this
handled by Stata in an appropriate manner?
(2) When I stset data with the command
stset year1, id(id) origin(time founding) enter(time start)
Here founding contains the time when the subject first came at risk
and start contains the time the observation window starts. So in a
way
I know when a company was founded, but I want to start analysis in a
certain year. Are these observations still left censored? Is this
handled by Stata?
Thanks for any help,
best wished,
dirk
*
* 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/
--
Dirk Volland
mobile +358468948277
skype junin7
! new Email: [email protected]
*
* 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/
*
* 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/