you may choose, depending on your data, between using a fixed effect model, a dynamic model (where lags can be used as instruments), or the inverse probability of treatment weights (see Epidemiology, 11: 550-560 and see Jamie Robins and Miguel Hernan both Harvard School of Public Health, they have a website with all the relevant papers).
Nicola
At 02.33 29/04/2008 -0400, "Luis Ortiz" wrote:
>Dear Statalisters,
>
>I am using discrete-time panel data for carrying out a survival analysis on
>the instant likelihood of finding a JOB MATCH; that is, a match between
>educational attainment and occupation. I'm afraid, though, that many of the
>covariates affecting the instant likelihood of occurrence of this event
>might be also affecting the entrance into the risk set. In other word, my
>independent variables might not be just explaining the occurrence of that
>event, but also the entrance into the population of risk, constituted by all
>those who are MISMATCHED.
>
>Put in this way, my research question would be actually the following:
>"Conditional on [a number of endogenous variables possibly affecting the
>entrance in a population of risk], which is the effect of some of them over
>the individual likelihood of leaving this state (job mismatch) and finding a
>job match"?.
>
>I suppose this endogeneity problem should be dealt with INSTRUMENTAL
>VARIABLES, and the corresponding commands using this approach in STATA (i.e.
>ivreg2 or xtivreg2), but finding a suitable, reliable IV is not always easy,
>and might just not be available in my data. Were this not feasible, as an
>approach, is there any other way of dealing with the endogeneity problem
>mentioned that could be easily implemented with STATA?
>
>Many thanks for your attention
>
>Luis Ortiz
*
* 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/
