dynparam (version 1.0.2)

distribution: Defining, serialising and printing distributions

Description

Distributions are used to define the domain of an integer_parameter() or a numeric_parameter().

Usage

distribution(lower, upper, ...)
distribution_function(dist)
quantile_function(dist)
# S3 method for distribution
as.list(x, ...)
as_distribution(li)
is_distribution(x)

Arguments

lower

Lower limit of the distribution.

upper

Upper limit of the distribution.

...

Fields to be saved in the distribution.

dist

A distribution object.

x

An object which might be a distribution.

li

A list to be converted into a distribution.

List of all currently implemented distributions

• expuniform_distribution()

• normal_distribution()

• uniform_distribution()

Serialisation

• as.list(dist): Converting a distribution to a list.

• as_distribution(li): Converting a list back to a distribution.

• is_distribution(x): Checking whether something is a distribution.

Defining a distribution

In order to create a new distribution named xxx, you need to create three functions.

• A xxx() function that calls distribution(...) %>% add_class("xxx") at the end.

• quantile_function.xxx(): The quantile function for converting between a uniform distribution and the xxx distribution.

• distribution_function.xxx(): The distribution function for converting between a uniform distribution and the xxx distribution.

Check the implementations of normal_distribution(), quantile_function.normal_distribution() and distribution_function.normal_distribution() for an example on how to do define these functions. Alternatively, check the examples below.

Details

See the sections below for more information each of the functions.

See Also

dynparam for an overview of all dynparam functionality.

Examples

# NOT RUN {
di <- uniform_distribution(lower = 1, upper = 10)
print(di)

li <- as.list(di)
di2 <- as_distribution(li)
print(di2)

# Defining a custom distribution, using the pbeta and qbeta functions
beta_distribution <- function(
  shape1,
  shape2,
  ncp,
  lower = -Inf,
  upper = Inf
) {
  di <- distribution(lower = lower, upper = upper, shape1, shape2, ncp)
  add_class(di, beta_distribution)
}

distribution_function.beta_distribution <- function(dist) {
  function(q) {
    stats::pbeta(q, shape1 = dist$shape1, shape2 = dist$shape2, ncp = dist$ncp)
  }
}

quantile_function.beta_distribution <- function(dist) {
  function(p) {
    stats::qbeta(p, shape1 = dist$shape1, shape2 = dist$shape2, ncp = dist$ncp)
  }
}
# }