chickadee » statistics » kde

Computes kernel density estimate for the given sequence using specified bandwidth (smoothing parameter) and kernel function. Returns a list of (x . density) pairs with n evaluation points (default 512).

Available kernels:

• 'gaussian or 'normal - Gaussian kernel (most common)
• 'epanechnikov - Epanechnikov kernel (most efficient)
• 'rectangular or 'uniform - Rectangular kernel
• 'triangular - Triangular kernel
• 'biweight or 'quartic - Biweight kernel

> (let* ((data '(1 2 3 4 5))
         (bw (kde-bandwidth-nrd data))
         (density (kde data bw 'gaussian 10)))
    (length density))
10
> (kde '(1 2 3) 1.0 'gaussian 5)
((-2.0 . 0.006649110640673518)
 (-0.75 . 0.07266464304354425)
 (0.5 . 0.25725941161130127)
 (1.75 . 0.2968430843637567)
 (3.0 . 0.1123450687407796))