define_dist: Define a multivariate beta (mbeta) distribution

In maxwestphal/SIMPle: Simultaneous Inference for Multiple Proportions

Define a multivariate beta (mbeta) distribution

Description

Define a multivariate beta (mbeta) distribution

Usage

define_dist(
  nu = 1,
  mean = 1/2,
  corr = 0,
  moments = NULL,
  gamma = NULL,
  vars = length(mean),
  varnames = NULL,
  groups = 1,
  groupnames = NULL,
  mode = c("auto", "full", "reduced"),
  tol = 0.01,
  project = TRUE,
  msg = TRUE
)

Arguments

nu	integer, 'pseudo-prior-observations'
mean	numeric of length vars, specifying the marginal means. Repeated if of length one.
corr	matrix, symetric, positive definite correlation matrix. If a single correlation
moments	matrix, symmetric matrix with mixed 2nd order moments of the target distribution. See details.
gamma	numeric vector, parameter of the underlying Dirichlet distribution. the base 2 logarithm of length(gamma) needs to be an integer and is the dimension of the resulting multivariate Beta distribution.
vars	integer, number of variables
varnames	character, variables names, needs to have length vars
groups	integer, number of groups (subpopulations) (default: 1)
groupnames	character, group names
mode	character, either "auto" (default), "full" or "reduced". See details.
tol	numeric, maximum relative deviation (default: 0.01) between specified and achieved moment matrix (or correlation matrix)
project	logical, project parameters to satisfy Frechet bounds (TRUE, default) or not (FALSE)
msg	logical, return messages (TRUE, default) or not (FALSE)

Details

Mode "full" implies that the full parameter vector gamma is used which results in nonzero mixed moments. In mode "reduced", a sparse representation with 1+m+(m-1)*m/2 parameters is used to model (prior) sample size, mean and covariance / correlation matrix. Per default (mode="auto"), the mode is set to "full" for dimension <= 10 and "reduced" for dimensions >10.

Value

A SIMPle.dist object representing the multivariate Beta distribution with given parameters.

Examples

define_dist(vars = 3, nu=10, mean=c(0.9,0.8,0.7), corr=0.5)

maxwestphal/SIMPle documentation built on April 11, 2024, 4:07 p.m.