3.5.2.17 Kernel2width

1 Definition:
- 1.1 When method = 0 (default)
- 1.2 When method = 1
2 Parameters:

Definition:

kernel2width( vX, vY, wx, wy[, int method=0, int grid=32]) returns the 2D kernel density optimal bandwidths (wx, wy) of X scale and Y scale for datasets (vX, vY) using two different methods.

When method = 0 (default)

Bivariate Kernel Density Estimator method is used.

This method offers bandwidth based on linear diffusion process.

When method = 1

Rule of Thumb method is used.

The estimation of wx and wy simply can be calculated by:

$w_x = \frac{\sigma_x}{2n^{1/6}}$

$w_y = \frac{\sigma_y}{2n^{1/6}}$

where n is the size of vector vX or vY, $\sigma_x$ is the sample standard variation for dataset vX, and $\sigma_y$ for dataset vY accordingly.

Parameters:

vX (input, vector): x values of distributed samples used to estimate bandwidth
vY (input, vector): y values of distributed samples used to estimate bandwidth
wx (output, double): output width for X scale, $w_x > 0$
wy (output, double): output width for Y scale, $w_y > 0$
method (input, int): method = 0 (default) for Bivariate Kernel Density Estimator method or 1 for Rule of Thumb method.
grid (input, int): input number of grids in X/Y direction for method=0, grid is a positive integer.