[6546] in s-news-athena
Re: Cumulative multivariate normal distribution.
daemon@ATHENA.MIT.EDU (Simon Jackman)
Wed Mar 1 20:20:36 1995
Date: Wed, 1 Mar 1995 18:55:44 -0600 (CST)
From: Simon Jackman <acbsjack@gosnell.spc.uchicago.edu>
To: Huiman Xie Barnhart <hxb@panda.sph.emory.edu>
Cc: s-news@utstat.toronto.edu
In-Reply-To: <Pine.ULT.3.90.950301160911.27694B-100000@panda.sph.emory.edu>
try the toms717 Fortran routine on statlib (under misc, I think).
This was contributed by Alan Miller (a fellow Aussie?!) from CSIRO, I
believe...
header is:
SUBROUTINE MECDF(NDIM, D, RHO, PROB, IER)
C
C This comes from TOMS algorithm 717
C
INTEGER NDIM, IER
DOUBLE PRECISION D(*), PROB, RHO(*)
C-----------------------------------------------------------------
C 6/29/90
C This subroutine is designed to calculate the MVN CDF
C using the Mendell-Elston procedure as described in
C Kamakura (1989). The current version is set up to go
C as high as 19 dimensions (=> 20 MNP alternatives)
C NOTE: Equation (15) in Kamakura has an error.
C
C Specifically, assume that Z is a set of random variables
C with a standard normal distribution with correlations
C stored in RHO (in packed form). Then this subroutine
C calculates Prob[Z(1)>D(1);...; Z(NDIM) > D(NDIM)].
C-----------------------------------------------------------------
I compiled this and put it local.Sqpe (static load only here, ver 3.1;
groan) call it with the following Splus function:
> mvncdf
function(x, rho)
{
# caller function for tom717.f - fortran routine for mecdf
x <- as.vector(x)
ndim <- length(x)
d <- x
# fortran call is
# SUBROUTINE MECDF(NDIM, D, RHO, PROB, IER)
out <- .Fortran("mecdf",
as.integer(ndim),
as.double(d),
as.double(rho),
as.double(1),
as.integer(1))
out
}
regards - simon jackman
========================================================================
Simon Jackman, Dept of Political Science, Univ of Chicago, 5828
S.University Ave, Chicago, IL 60637. ph: 312.702.8075. fax: x.x.1689 "But
the age of chivalry is gone. That of sophisters, economists, and
calculators has succeeded; and the glory of Europe is extinguished for
ever." (Burke, _Reflections on the Revolution in France_) "I can't give
credit enough to the logisticians..." (Norman Schwarzkopf, 2/27/91)
On Wed, 1 Mar 1995, Huiman Xie Barnhart wrote:
>
> I am looking for some functions in S-plus which compute the
> cumulative bivariate or multivariate normal distribution. Any suggestions
> will be greatly appreciated.
>
> -----------------------------------------------------------------------
> Huiman X. Barnhart, Ph.D. //\ /\\
> Assistant Professor
> Department of Biostatistics || * \ . . / * ||
> Emory School of Public Health \\____\X/____//
> e-mail: hxb@panda.sph.emory.edu
> (404)727-8059 / * /O\ * \
> (404)727-1370(fax) \__/ " \__/
>
>
>
