st: Assign year to observations (failures) based on proportion of successes in each year.

[Brad Fedy <bcfbradley@gmail.com>]

I am interested in conducting a logistic regression with success=1 and
failure =0.  My success cases have a specific year associated with
them - my failures do not.  I have annual columns for each covariate
that are structured as: x_year, e.g. x_1998, x_1999, x1_1998, x1_1999.
 I have extracted the correct annual covariate value for the successes
(1) using code similar to this:

gen x_correct=.
forval i=1998/2008 {
    replace x_correct if x_year == `i' & success==1
    }

I have many more failures than successes.  I want to assign a
particular year to each of the failures.  I want to distribute the
assignment of years to the failures based on the proportion of
successes that fell in a particular year. For example: if 25% of the
successful observations were in 1998, and 75% in 1999 I want to assign
a year value of 1998 to 25% of the failures and 1999 to the remaining
75% of the failures.  The data is panel structured, and therefore I
have to do this across multiple grouping variables e.g. households.

I would really appreciate any suggestions.

Thanks,

*
*   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/