st: unbalanced panel estimates with lags and dep var censoring

[Lloyd Dumont <lloyddumont@yahoo.com>]

Hello, everyone.

I am facing what I’m sure is a fairly typical data-related issue, and I am trying to figure out the best way to deal with it.

I have an unbalanced panel of units within an organization.  Units were chosen at random for a survey once per year.  So, over a 7 year period, each unit appears anywhere from once to—in a few cases—four times.  The data are xtset, as you might expect.

As I wrap my head around this, here is the key issue (though more come up as I think about this).  Let’s say the key dependent variable only exists from year 4 onward.  If you just run a regular xtreg, it will drop all observations pre-year 4.  But, lagged values on some independent variables could be useful.  Adding lags is not difficult, but there will need to be a whole vector of lagged variables, since each observation comes and goes.  And, unless a unit is present for all years, which it never is, then it will drop out of the estimates altogether.

So, I want to be able to estimate a dep var using lagged indep vars, when they exist.  And, the model should probably account for the fact that some observations come from YEAR 1, some from YEAR 2, etc.  

Does that make any sense?  I would be happy to provide more detail.

Thank you for your thoughts.  Lloyd


      

*
*   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/