| kernel.pow | R Documentation |
Builds the name of a kernel of the power family
K_{s,t}(x) = c_{s,t}\,(1-|x|^s)^t\,\mathbf{1}_{|x|\le 1},
\qquad c_{s,t} = \frac{s}{2\,B(1/s,\,t+1)},
which can be passed as kernel to every function of the package
(bandwidth selectors, dkde, kernel.fun,
kernel.conv).
kernel.pow(s = 2, t = 2)
s |
single positive number: shape exponent |
t |
single non-negative number: power |
The kernel can equally be given directly as the string "pow(s,t)",
e.g. kernel = "pow(2,3)"; "pow" alone means "pow(2,2)".
The family contains several classical kernels:
"pow(s,0)" | uniform |
"pow(1,1)" | triangular |
"pow(2,1)" | Epanechnikov |
"pow(2,2)" | biweight |
"pow(2,3)" | triweight |
"pow(3,3)" | tricube |
The constants used by the bandwidth selectors are exact:
\mu_k(K) = B((k+1)/s,\,t+1)/B(1/s,\,t+1) for even k and
R(K) = s\,B(1/s,\,2t+1) / (2\,B(1/s,\,t+1)^2). For x > 0,
K^{(r)}(x) is a finite sum of terms
a\,x^{p}(1-x^s)^{q}: an exact binomial expansion when t is an
integer (t \le 10), a recurrence otherwise; derivatives are available
up to order 16.
R(K^{(r)}) is finite only when 2r < 2t+1 and, for a non-integer
s, 2r < 2s+1 (otherwise K^{(r)} is not square integrable
at |x| = 1 or at 0). Bandwidth selectors return NA with a
warning for such derivative orders, as for the other finite-support kernels.
Maximum derivative order. K^{(r)} \ne 0 up to
r = \min(st, 16) when s and t are
integers (16 otherwise) in kernel.fun,
kernel.conv and dkde. With r^* the
largest integer r < t + 1/2 (with 2r \le st for integer
s and t, and r < s + 1/2 unless
s is an even integer), the maximum is \min(r^*, 6) for
h.ucv, h.bcv, h.tcv and
h.amise, \min(r^*-1, 6) for
h.mcv and \min(r^*-2, 6) for
h.ccv. For example pow(2,3): 6, 3, 2, 1;
pow(2,8): 16, 6, 6, 6; pow(1.5,2.5): 16, 1, 0, not available.
A character string "pow(s,t)".
Arsalane Chouaib Guidoum acguidoum@univ-tam.dz, Amel Henider amel.henider@univ-tam.dz and Iago Giné-Vázquez iago.gin-vaz@protonmail.com
kernel.fun, kernel.conv, h.ucv,
dkde.
kernel.pow(2, 3) # "pow(2,3)"
x <- rnorm(200)
h.ucv(x, kernel = kernel.pow(2, 2)) # same bandwidth as kernel = "biweight"
h.ucv(x, kernel = "pow(1.5,2.5)")
plot(kernel.fun(kernel = kernel.pow(1.5, 2.5), deriv.order = 1L))
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.