Skip to content

Computing inverse cdf (quantile function) from a KDE

3 messages · firdaus.janoos, David L Carlson, Greg Snow

David L Carlson
# 
Something like this? 

library(KernSmooth)
x <- rnorm(100)
KS <- bkde(x)
diff <- KS$x[2] - KS$x[1]
yc <- cumsum(KS$y*diff)
approx(yc, KS$x, runif(1000))

If you are going to use approx() repeatedly, you can create a
function with approxfun().

-------------------------------------
David L Carlson
Associate Professor of Anthropology
Texas A&M University
College Station, TX 77840-4352


----- Original Message ----- 

From: "firdaus.janoos" <fjanoos at bwh.harvard.edu> 
To: r-help at r-project.org 
Sent: Wednesday, July 11, 2012 9:33:30 AM 
Subject: [R] Computing inverse cdf (quantile function) from a KDE 

Hello, 

I wanted to know if there is a simple way of getting the inverse cdf for a 
KDE estimate of a density (using the ks or KernSmooth packages) in R ? 

The method I'm using now is to perform a numerical integration of the pdf 
to get the cdf and then doing a search for the desired probablity value, 
which is highly inefficient and very slow. 

Thanks, 
-fj 


______________________________________________ 
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.
Greg Snow
# 
If you are going to be doing a lot of this then you might want to
consider using logspline density estimates (logspline package) instead
of kernel density estimates.
On Wed, Jul 11, 2012 at 8:33 AM, firdaus.janoos <fjanoos at bwh.harvard.edu> wrote: