iprior: Imprecise Prior Specification

In ipeglim: Imprecise Inferential Framework on Statistical Reasoning

Description

The function name iprior stands for imprecise prior. A set of $k$ linear inequality constraints is used for modelling iprior. The details of this specification are given under ‘Details’.

Usage

  iprior(obj, eqns = list(), circle = list(), x, mat,
    verbose = FALSE, ...)

  ## Default S3 method:
 iprior(obj, eqns = list(),
    circle = list(), x, mat, verbose = FALSE, ...)

  ## S3 method for class 'imprecise'
 iprior(obj, eqns = list(),
    circle = list(), x, mat, verbose = FALSE, ...)

Arguments

obj	an object of class imprecise produced from model.
eqns	a list with components of lhs and rhs; See ‘Details’ for more information.
circle	a list with components of x, y, z, r, and len. See ‘Details’ for more information.
x	a numeric vector.
mat	a numeric matrix.
verbose	a logical value; Be more verbose about the process by displaying messages. Defaults to FALSE.

Details

Modelling a prior ignorance is the second stage on the imprecise inferential framework. stage has the value "iprior" that is the name of environment called.

A convex hull is defined by lhs %*% x >= rhs, where lhs is a constraint matrix k x p and rhs is a constraint vector of length k from k linear inequality constraints. The function iprior searches for the solutions xtms of this system which are the extreme points of that convex hull.

Depending on the numbers of dimensions p and constraints k, a shape of this convex hull is determined. For example, a box-constrained convex hull is made on p=2 dimensional space with k=4 constraints. When k has an infinite number, a nutural shape defaults to a circle on p=2 dimensional space.

Value

An object of class imprecise with the components:

stage	the environment name called.
xtms	the list of solutions of the system supplied.

Note

Conventionally, the current version classifies a shape of convex hull into five categories: eqns2d, eqns3d, eqns4more, circle2d, and sphere3d for visualization.

It is a difficult task to sarch for extreme points in more than three dimensional cases. The function convhulln in the package geometry is a good replacement of iprior. (See more details about convhulln at http://geometry.r-forge.r-project.org/).

Author(s)

Chel Hee Lee <[e-mail]>

References

Lee (2013) “Imprecise inferential framework”, PhD thesis.

See Also

chull, convhulln,

Examples

## Not run: 
lc0 <- list(lhs=rbind(diag(2), -diag(2)), rhs=c(rep(-1,4)))
cmfit <- iprior(obj=m2fit, eqns=lc0)
plot(cmfit)

## End(Not run)

ipeglim documentation built on May 2, 2019, 4:31 p.m.