[R] Non-crossing Nonparametric quantile regressions

[roger koenker]

On Nov 11, 2007, at 2:20 PM, paulbegc wrote:
>
> I've been looking for ways to calculate a large number (100) of non- 
> crossing
> Nonparametric quantile regressions on large populations (1000+).
>
> Can the quantreg package in R ensure the non-crossing property?

You can estimate a (reasonably) large number of quantiles simultaneously
and constrain them to be monotone at a (reasonably) large number of  
points
but this is rather messy, and also pushes the estimation back toward  
a almost
parametric specification.  I prefer to let quantiles to be free --  
determined by
only local information, not constrained by a global specification.
>
> If not, do you know any alternative?

Another alternative is to estimate the quantile regressions without  
constraints and
then use some monotonization technique like pool adjacent violators or
rearrangement to enforce monotonicity at various x-points.  See

	?rearrange

in the quantreg package for an example of how to use the latter approach
and for references.


url:    www.econ.uiuc.edu/~roger                Roger Koenker
email   rkoenker at uiuc.edu                       Department of Economics
vox:    217-333-4558                            University of Illinois
fax:    217-244-6678                            Champaign, IL 61820

>
> Thank you,
>
> Paulbegc
> -- 
> View this message in context: http://www.nabble.com/Non-crossing- 
> Nonparametric-quantile-regressions-tf4787333.html#a13695430
> Sent from the R help mailing list archive at Nabble.com.
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting- 
> guide.html
> and provide commented, minimal, self-contained, reproducible code.