st: Discrete-time duration models with sample selection

["Pattaro S." <S.Pattaro@soton.ac.uk>]

Hello,

I was wondering whether members of Statalist know of the existence of a Stata programme or utility that can deal with sample selection within the context of discrete-time duration models. In the literature it is common to find applications based on heckprob models for the estimation of Heckman two-step models for a binary outcome. I haven't come across any application that use these types of models (bivariate probit with heckman selection) within a discrete-time context.

The substantive example I am currently working on relates to a British cohort born in 1970 who has been followed up since birth to age 38. Detailed information about respondents and their partners at ages 16, 30, 34 and 38 are available. The model takes those who are childless at age 30  and in a couple and looks at the factors affecting transitions to first birth between 30 and 38. Of course, those who remain childless at age 30 are a select group and so in our selection model we would like to perform a probit (or logit) of remaining childless at 30 and then a discrete time hazards model of progression to first conception between age 30 and 38. A mixture of cross-sectional variables (measured at different survey time points) and time-varying variables (measured on a monthly basis) are available.

In this particular case, cross-sectional variables are entered in the selection equation while the second equation -to estimate the outcome of interest- is based on both fixed- and time-varying parameters.

Could anyone help on this matter and point me to a Stata utility that could be applied in this specific case or provide any reference or empirical application within the context of discrete-time hazards models?

Many thanks in advance.

Best,
Serena
*
*   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/