gbker1
Produce kernel density estimation
Usage
gbker [options]
Overview
Compute the kernel density estimation on an equispaced grid. Data are read from standard input. The kernel bandwidth, if not provided with the option -H, is set automatically using simple heuristics. Normally the whole support is displayed. The option -w can be used to restrict the computation to specific range of values. The method used to compute the density is set by the option -M. The boundaries of the grid are slightly different in the different cases.
Options
-
-n -
number of equispaced points/bins where the density is computed (default 64)
-
-w -
set the grid boundaries with min,max (default whole support)
-
-H -
set the kernel bandwidth
-
-S -
scale the automatic kernel bandwidth
-
-K -
choose the kernel to use (default 0)
0
|
Epanenchnikov |
|---|---|
1
|
Rectangular |
2
|
Gaussian |
3
|
Laplacian |
-
-M -
choose the method for density computation (default 0)
0
|
use FFT (# grid points rounded to nearest power of 2) |
|---|---|
1
|
use discrete convolution (compact kernels only; bins>2) |
2
|
explicit summation (boundaries excluded) |
-
-F -
specify the input fields separators (default " \t")
Examples
- gbker -n 128 -K 2 < file
-
use a Gaussian kernel to compute the density on 128 equispaced points with the FFT approach. All the data in 'file' are used, multiple columns are pooled.
- gbker -K 3 -M 2 < file
-
use a Laplacian kernel to compute the density on 64 points. The method used is the explicit summation.
Releases
5 releases since Debian 9 (Stretch); the text changed 3 times.