st: logistic modeling with longitudinal data (NLSY)

[Matthew Aronson <maronson@rams.colostate.edu>]

not address any specific Stata issue.] To wit: I'm using a logistic
regression model to examine factors related to whether a person has
ever completed high school. (The broader purpose is to compare this
binary approach with an approach that treats high school completion as
a time-varying outcome in a survival analysis.) I'm thinking to
include a term in the model that reflects how many "years beyond
ordinary high school graduation age" each respondent was observed, but
I'm not sure of the right way to define that term.

My data derives from the NLSY, a U.S. study in which a sample of
respondents between age 14 and 22 all entered the study in 1979, and
all of them were followed up until 2004.  Due to loss to follow up and
different ages at entry, some subjects might have been followed up to
age 21, and therefore had about 3 years past "ordinary high school
graduation age" to have been observed to have ever graduated, while
others might have been followed up to age 35, and thus had about 17
years to have been observed to have ever graduated.

My first inclination was to simply define a "years of observation
beyond 18" using the very last (latest) age at which the respondent
was interviewed.  Thus, a person who was observed up to age 30 would
have a value of 12, whether or not s/he ever graduated from high
school.  But, on the other hand, I had the thought that once a person
graduates from high school, s/he is not in the risk set, so perhaps
the years of observation variable should be given the age at high
school graduation, so that, e.g., a  person observed up to age 30 but
who graduated from high school at age 19 would be given a value of 1.

I wonder if anyone can offer some guidance here, thanks very much.
Matt Aronson
*
*   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/