Re: Quantile regression - computation of differences

[Brian S Cade <brian_cade@usgs.gov>]

Patrick:  If I'm understanding
your issue correctly, it sounds like you are separately estimating quantile
regressions for males and for females.  To get the estimated differences
in coefficients by quantiles (and CI), you need to set this up as one model
for both sexes by including an indicator variable for gender and the interaction
between the indicator variable and the other predictor(s).   Then
when you estimate the model, your intercept and predictor variable estimates
will be the values for one gender (male or female, depending on how you
code the indicator variable); and the estimates for the indicator variable
and its interactions will be the differences between the genders for the
intercept and predictor(s), respectively.   This is just what would
be done with least square regression, e.g., see Neter et al. (1996: 468-473)
for comparison of two or more regression functions.   Since the differences
between genders are estimated directly, you should be able to get plot
of estimates by quantiles with their confidence bands. 

Brian

Brian S. Cade

U. S. Geological Survey
Fort Collins Science Center
2150 Centre Ave., Bldg. C
Fort Collins, CO  80526-8818

email:  brian_cade@usgs.gov
tel:  970 226-9326

<Patrick.Heer@bakom.admin.ch>
Sent by: s-news-owner@lists.biostat.wustl.edu

04/27/2006 12:26 AM

To	<s-news@lists.biostat.wustl.edu>
cc
Subject	[S] Quantile regression - computation of differences

Dear all
I am using R 2.2.1 with the quantreg package of Koenker.
I am trying to compute and plot the differences between 2 quantile
regression models (one for male and the other one for female) for
several taus.
I can easily compute the results for the coefficients but how can plot
the results in the same way as plot(summary(x, se="...") so that
I can
see the confidence band.
Thanks for your help

Patrick

--------------------------------------------------------------------
This message was distributed by s-news@lists.biostat.wustl.edu.  To
unsubscribe send e-mail to s-news-request@lists.biostat.wustl.edu with
the BODY of the message:  unsubscribe s-news