I imagine there is a connection here to Martin's earlier post.
Given
regress <response> <indicator>
you can recover the means for indicator == 0 and indicator == 1. That's
part of the magic of least squares.
But it's not necessarily true that given
qreg <response> <indicator>
you can recover the equivalent medians. Working with a norm based on
absolute differences doesn't have that side effect in general.
Nick
[email protected]
Newson, Roger B
The median difference is not the same thing as a difference between
medians. This distinction causes a lot of culture shock to people who
are accustomed to working with means and their differences.
*
* 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/
