nimble (version 1.3.0)

StickBreakingFunction: The Stick Breaking Function

Description

Computes probabilities based on stick breaking construction.

Usage

stick_breaking(z, log = 0)

Arguments

z

vector argument.

log

logical; if TRUE, weights are returned on the log scale.

Author

Claudia Wehrhahn

Details

The stick breaking function produces a vector of probabilities that add up to one, based on a series of individual probabilities in z, which define the breaking points relative to the remaining stick length. The first element of z determines the first probability based on breaking a proportion z[1] from a stick of length one. The second element of z determines the second probability based on breaking a proportion z[2] from the remaining stick (of length 1-z[1]), and so forth. Each element of z should be in \((0,1)\). The returned vector has length equal to the length of z plus 1. If z[k] is equal to 1 for any k, then the returned vector has length smaller than z. If one of the components is smaller than 0 or greater than 1, NaNs are returned.

References

Sethuraman, J. (1994). A constructive definition of Dirichlet priors. Statistica Sinica, 639-650.

Examples

z <- rbeta(5, 1, 1)
stick_breaking(z)

if (FALSE) {
cstick_breaking <- compileNimble(stick_breaking)
cstick_breaking(z)
}