sns.make.part {sns}

Utility Functions for Creating and Validating State Space Partitions

Description

Utility functions for creating and validating state space partitions, to be used in SNS for improving the mixing of sampled chains for high-dimensional posteriors.

Usage

sns.make.part(K, nsubset, method = "naive")
sns.check.part(part, K)

Arguments

K

Dimensionality of state space.

nsubset

Number of subsets to partition the state space dimensions into.

method

Method used for state space partitioning. Currently, only naive method is implemented, where coordinates are distributed evenly (or as evenly as possible) across subsets.

part

A list of length nsubset, with each element a vector of integer values, representing the coordinates belonging to a subset. This list is the output of sns.make.part.

Value

sns.make.part produces a list of integer vectors, each containing coordinates belonging to the same subset. sns.check.part produces a boolean flag, indicating whether or not the partition list is valid or not. The subset members must constitute a mutually-exclusive, collectively-exhaustive set relative to 1:K.

Author(s)

Alireza S. Mahani, Asad Hasan, Marshall Jiang, Mansour T.A. Sharabiani

References

Mahani A.S., Hasan A., Jiang M. & Sharabiani M.T.A. (2016). Stochastic Newton Sampler: The R Package sns. Journal of Statistical Software, Code Snippets, 74(2), 1-33. doi:10.18637/jss.v074.c02

See Also

sns, sns.run

Examples

# creating a valid partition of a 6-dimensional state space
my.part.valid <- list(c(1,2,3), c(4,5,6))
is.valid.1 <- sns.check.part(my.part.valid, 6)
cat("is partition valid: ", is.valid.1, "\n")

# creating an invalid partition of a 6-dimensional state space
# (coordinate 4 is missing)
my.part.invalid <- list(c(1,2,3), c(5,6))
is.valid.2 <- sns.check.part(my.part.invalid, 6)
cat("is partition valid: ", is.valid.2, "\n")