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
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
|
\name{interface}
\alias{mvnorm}
\alias{aperm.mvnorm}
\alias{simulate.mvnorm}
\alias{logLik.mvnorm}
\alias{lLgrad}
\alias{lLgrad.mvnorm}
\alias{margDist}
\alias{margDist.mvnorm}
\alias{condDist}
\alias{condDist.mvnorm}
\title{
(Experimental) User Interface to Multiple Multivariate Normal Distributions
}
\description{
A (still experimental) simple user interface for computing on multiple multivariate
normal distributions.
}
\usage{
mvnorm(mean, chol, invchol)
\S3method{aperm}{mvnorm}(a, perm, ...)
margDist(object, which, ...)
\S3method{margDist}{mvnorm}(object, which, ...)
condDist(object, which_given, given, ...)
\S3method{condDist}{mvnorm}(object, which_given, given, ...)
\S3method{simulate}{mvnorm}(object, nsim = dim(object$scale)[1L], seed = NULL,
standardize = FALSE, as.data.frame = FALSE, ...)
\S3method{logLik}{mvnorm}(object, obs, lower, upper, standardize = FALSE, ...)
\S3method{lLgrad}{mvnorm}(object, obs, lower, upper, standardize = FALSE, ...)
}
\arguments{
\item{chol}{either an \code{ltMatrices} object specifying (multiple)
Cholesky factors of the covariance matrix or
one single numeric lower triangular square matrix.
}
\item{invchol}{either an \code{ltMatrices} object specifying (multiple)
inverse Cholesky factors of the covariance matrix or
one single numeric lower triangular square matrix.
}
\item{a,object}{objects of class \code{mvnorm}.
}
\item{perm}{a permutation of the covariance matrix corresponding to \code{a}.
}
\item{which}{names or indices of elements those marginal distribution
is of interest.
}
\item{which_given}{names or indices of elements to condition on.
}
\item{given}{matrix of realisations to condition on (number of rows is
equal to \code{length(which)}, the number of
columns corresponds to the number of matrices in \code{chol}
or \code{invchol}.
}
\item{lower}{matrix of lower limits (one column for each observation, \eqn{J} rows).
}
\item{upper}{matrix of upper limits (one column for each observation, \eqn{J} rows).
}
\item{obs}{matrix of exact observations (one column for each observation, \eqn{J} rows).
}
\item{mean}{matrix of means (one column for each observation, length is
recycled to length of \code{obs}, \code{lower} and \code{upper}).
}
\item{seed}{an object specifying if and how the random number generator
should be initialized, see \code{\link[stats]{simulate}}.
}
\item{standardize}{logical, should the Cholesky factor (or its inverse) undergo
standardization (ensuring the covariance matrix is a correlation
matrix) before computing the likelihood.
}
\item{nsim}{number of samples to draw.
}
\item{as.data.frame}{logical, convert the $J x N$ matrix result to a
classical $N x J$ data frame.
}
\item{\dots}{Additional arguments to \code{\link{ldpmvnorm}} and
\code{\link{sldpmvnorm}}
}
}
\details{
The constructor \code{mvnorm} can be used to specify (multiple)
multivariate normal distributions. \code{margDist} derives marginal and
\code{condDist} conditional distributions from such objects. A
\code{simulate} method exists for drawn samples from multivariate
normals.
The continuous (data in \code{obs}), discrete (intervals in \code{lower}
and \code{upper}), and mixed continuous-discrete log-likelihood is
implemented in \code{logLik}. The corresponding gradients with respect
to all model parameters and with respect to the data arguments
is available from \code{lLgrad}.
Rationals and examples are given in Chapter 7 of the package vignette
linked to below.
}
\value{
\code{mvnorm}, \code{margDist}, and \code{condDist} return objects
of class \code{mvnorm}. \code{logLik} returns the log-likelihood
and \code{lLgrad} a list with gradients.
}
\seealso{\code{vignette("lmvnorm_src", package = "mvtnorm")}}
\keyword{distribution}
|
