On Jun 27, 2011, at 3:18 PM, Carsten Harlaß wrote:
Hello,
thanks for your response.
Of course I already thought about this "simple" solution of the
problem.
But I think this is not a nice workaround.
If I understand the manual correctly, density() already makes an
approximation. But this approximation is just evaluated at the wrong
points.
See:
"Details
The algorithm used in density.default disperses the mass of the
empirical distribution function over a regular grid of at least 512
points and then uses the fast Fourier transform to convolve this
approximation with a discretized version of the kernel and then uses
linear approximation to evaluate the density at the specified points."
and
"n the number of equally spaced points at which the density is to be
estimated. When n > 512, it is rounded up to a power of 2 during the
calculations (as fft is used) and the final result is interpolated by
approx. So it almost always makes sense to specify n as a power of
two. "
So this workaround means putting a second approximation on top of an
other approximation. That's not nice. Or is it?
It's just a linear approximation between two points of the density$.
If you need to you can make those point more numerous. What
alternative are you considering?
(I would also point out that approxfun may be even more helpful. It
gives you the result as a function.)
--
David
With kind regards
Carsten
Am 27.06.2011 19:16, schrieb David L Carlson:
Look at ?approx. For your example (of course your random numbers give
different results):
approx(f$x, f$y, c(-2, -1, 0, 1, 2))
$x
[1] -2 -1 0 1 2
$y
[1] 0.03757113 0.19007982 0.31941779 0.37066592 0.10227509
approx gives NA's if you try to interpolate outside the bounds of
the data.
----------------------------------------------
David L Carlson
Associate Professor of Anthropology
Texas A&M University
College Station, TX 77843-4352
-----Original Message-----
From: r-help-boun...@r-project.org [mailto:r-help-boun...@r-project.org
] On
Behalf Of Carsten Harlaß
Sent: Sunday, June 26, 2011 7:02 PM
To: r-help@r-project.org
Subject: [R] Kernel Density Estimation at manually specified points
Hello,
my name is Carsten. This ist my first post to R-help mailing list.
I estimate densities with the function "density" out of the package
"stats".
A simplified example:
#generation of test data
n=10
z = rnorm(n)
#density estimation
f=density(z,kernel="epanechnikov",n=n)
#evaluation
print(f$y[5])
Here I can only evaluate the estimation at given points. These points
are determined by the parameter n. By default they are equidistant
distributed on the interesting interval.
But I need to evaluate the estimation (the estimated densitiy
function)
at manually specified points. For example I want to compute f(z[i]).
This means I am interested in the estimated density at a the
observation
z[i].
Does anyone know how I can compute this? I think this is an ordinary
task so I would be surprised if R can not manage this. But even
after a
long search I have found nothing.
Thanks in advance
Carsten Harlaß
--
Carsten Harlaß
Aachen University of Applied Sciences
Campus Jülich
E-Mail: carsten_harl...@web.de
______________________________________________
R-help@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.
--
Carsten Harlass
Aachen University of Applied Sciences
Campus Juelich
E-Mail: carsten_harl...@web.de
______________________________________________
R-help@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.
David Winsemius, MD
West Hartford, CT
______________________________________________
R-help@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.
