Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | <andri.koch@iop.unibe.ch> |
To | <statalist@hsphsun2.harvard.edu> |
Subject | AW: st: stset survival analysis with right censoring and left truncation for a bankruptcy dataset |
Date | Tue, 31 Dec 2013 12:26:33 +0000 |
Thanks Steve for your help. So I'll consider logit and log-log models in more detail. Andri ________________________________________ Von: owner-statalist@hsphsun2.harvard.edu [owner-statalist@hsphsun2.harvard.edu]" im Auftrag von "Steve Samuels [sjsamuels@gmail.com] Gesendet: Sonntag, 29. Dezember 2013 17:25 An: statalist@hsphsun2.harvard.edu Betreff: Re: st: stset survival analysis with right censoring and left truncation for a bankruptcy dataset Your analysis is attempting to predict firm bankruptcy from a data set of workers. This won't work. For a bankrupt firm with 500 workers, for example,the -stset- statement would tell Stata that there were 500 failures. You must create a data set with one observation per firm per year and create variables that summarize worker characteristics in the prior year; one variable might be proportion female. I say "prior year", because an observed association in current year data might arise because a bankruptcy led to changes in the worker mix. Also: Stata's -st- commands require failures in a time dimension with enough values that it can be considered continuous. With only yearly data, you must use a discrete analysis. I recommend either a logit or complementary log-log model (-cloglog-). For more details, see the the Lesson 6 link to discrete data analysis on Stephen Jenkins's fine web page "Survival analysis with Stata" (http://www.iser.essex.ac.uk/survival-analysis) Steve Samuels sjsamuels@gmail.com 18 Cantine's Island Saugerties NY 12477 USA 845-246-0774 > On Dec 28, 2013, at 9:02 AM, <andri.koch@iop.unibe.ch> <andri.koch@iop.unibe.ch> wrote: > > Dear all, > > I have a panel dataset over 5 years with the following variables: pers_id, employee, company, entry, exit, firm's bankruptcy_year, status, outcome. Where: > pers_id=identification number of employee i in company j. > employee=identification number of employee i. > company=identification number of company j. > entry/exit: start and end of the labor contract for employee i. > firm's bankruptcy_year: no explanation needed > status: 0 when the company survived, 1 when became insolvent > outcome: 1 if failed, 0 the observation was not observed to fail. > > Severall data are right censored and or left truncated. > There are much more variables in the dataset, i.e. age, sex, function.... > > I want to run a survival analysis to observe the employee's influence on a bankruptcy. Some people are still employed at the time of the bankruptcy but some people have left the company before the bankruptcy. I think that the former people are well considered with command: > > stset end, id(pers_id) time0(start) origin(time start) failure(outcome==1) > > but with that command the latter group of people aren't correct involved. Because the may have influenced the bankruptcy too. > > I hope that someone could advice me how to involve the latter group correct in my stset. > > Regards, > Andri > * * 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/ * * 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/