st: taking out the correct year-fixed-effects.

[andreas nordset <andreas.nordset@gmail.com>]

Dear Listers,

I have a balanced panel of households from 1995 to 2007.

They are all hit by an unemployment shock in either of the years 1999-2003.
Hence I can observe ALL households as early as 4 years prior to and 4
years after the shock.

I run the regression

-reg income relative-year-dummies calendar-year-dummies-

,where i leave out the calendar year dummy for 2004, to make that as a
"base year".
When i then plot the Constant+ relative year dummies in a diagram from
relative year -4 to +4 where Year-Fixed effects are washed out , I can
roughly state that "Income values are in 2004-$".

Maarten simplified this work significantly by suggesting to run something like:

-reg income relative-year-dummies calendar-year-dummies-
-adjust calendar-year-dummies, by(relative_year)-

This gives me a nice table with correct values for each of the
relative years. However, this procedure gives the plot a different
level. So I can no longer state that values are in the value of some
base year. Does anyone know a way to set it up so I can again
interpret the values to be of 2004-units ( or any other year).


Thanks a lot for the help,

Andreas
*
*   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/