1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
|
\name{gauss.quad.prob}
\alias{gauss.quad.prob}
\title{Gaussian Quadrature with Probability Distributions}
\description{Calculate nodes and weights for Gaussian quadrature in terms of probability distributions.}
\usage{
gauss.quad.prob(n, dist = "uniform", l = 0, u = 1,
mu = 0, sigma = 1, alpha = 1, beta = 1)
}
\arguments{
\item{n}{number of nodes and weights}
\item{dist}{distribution that Gaussian quadrature is based on, one of \code{"uniform"}, \code{"normal"}, \code{"beta"} or \code{"gamma"}}
\item{l}{lower limit of uniform distribution}
\item{u}{upper limit of uniform distribution}
\item{mu}{mean of normal distribution}
\item{sigma}{standard deviation of normal distribution}
\item{alpha}{positive shape parameter for gamma distribution or first shape parameter for beta distribution}
\item{beta}{positive scale parameter for gamma distribution or second shape parameter for beta distribution}
}
\value{A list containing the components
\item{nodes}{vector of values at which to evaluate the function}
\item{weights}{vector of weights to give the function values}
}
\details{
This is a rewriting and simplification of \code{gauss.quad} in terms of probability distributions.
The probability interpretation is explained by Smyth (1998).
For details on the underlying quadrature rules, see \code{\link{gauss.quad}}.
The expected value of \code{f(X)} is approximated by \code{sum(w*f(x))} where \code{x} is the vector of nodes and \code{w} is the vector of weights. The approximation is exact if \code{f(x)} is a polynomial of order no more than \code{2n-1}.
The possible choices for the distribution of \code{X} are as follows:
Uniform on \code{(l,u)}.
Normal with mean \code{mu} and standard deviation \code{sigma}.
Beta with density \code{x^(alpha-1)*(1-x)^(beta-1)/B(alpha,beta)} on \code{(0,1)}.
Gamma with density \code{x^(alpha-1)*exp(-x/beta)/beta^alpha/gamma(alpha)}.
}
\references{
Smyth, G. K. (1998). Polynomial approximation.
In: \emph{Encyclopedia of Biostatistics}, P. Armitage and T. Colton (eds.), Wiley, London, pages 3425-3429.
\url{http://www.statsci.org/smyth/pubs/PolyApprox-Preprint.pdf}
}
\author{Gordon Smyth, using Netlib Fortran code \url{https://netlib.org/go/gaussq.f}, and including corrections suggested by Spencer Graves}
\seealso{
\code{\link{gauss.quad}}, \code{\link{integrate}}
}
\examples{
# the 4th moment of the standard normal is 3
out <- gauss.quad.prob(10,"normal")
sum(out$weights * out$nodes^4)
# the expected value of log(X) where X is gamma is digamma(alpha)
out <- gauss.quad.prob(32,"gamma",alpha=5)
sum(out$weights * log(out$nodes))
}
\keyword{math}
|
