nint_tanTransform: Tangent Transform

In docopulae: Optimal Designs for Copula Models

Description

nint_tanTransform creates the transformation g(x) = atan((x - center)/scale) to be used in nint_transform.

Usage

nint_tanTransform(center, scale, dIdcs = NULL)

Arguments

center, scale	see g(x).
dIdcs	an integer vector of indices, the dimensions to transform.

Value

nint_tanTransform returns a named list of two functions "g" and "giDgi" as required by nint_transform.

See Also

nint_transform

Examples

mu = 1e0
sigma = mu/3
f = function(x) dnorm(x, mean=mu, sd=sigma)
space = nint_space(nint_intvDim(-Inf, Inf))

tt = nint_transform(f, space, list(nint_tanTransform(0, 1, dIdcs=1)))
tt$space
ff = Vectorize(tt$f); curve(ff(x), tt$space[[1]][1], tt$space[[1]][2])

nint_integrate(tt$f, tt$space) # should return 1

# same with larger mu
mu = 1e4
sigma = mu/3
f = function(x) dnorm(x, mean=mu, sd=sigma)

tt = nint_transform(f, space, list(nint_tanTransform(0, 1, dIdcs=1)))
ff = Vectorize(tt$f); curve(ff(x), tt$space[[1]][1], tt$space[[1]][2])

try(nint_integrate(tt$f, tt$space)) # integral is probably divergent

# same with different transformation
tt = nint_transform(f, space, list(nint_tanTransform(mu, sigma, dIdcs=1)))
ff = Vectorize(tt$f); curve(ff(x), tt$space[[1]][1], tt$space[[1]][2])

nint_integrate(tt$f, tt$space) # should return 1

docopulae documentation built on May 2, 2019, 7:53 a.m.