Learn R Programming

sams (version 0.4.3)

Merge-Split Samplers for Conjugate Bayesian Nonparametric Models

Description

Markov chain Monte Carlo samplers for posterior simulations of conjugate Bayesian nonparametric mixture models. Functionality is provided for Gibbs sampling as in Algorithm 3 of Neal (2000) , restricted Gibbs merge-split sampling as described in Jain & Neal (2004) , and sequentially-allocated merge-split sampling , as well as summary and utility functions.

Copy Link

Version

Install

install.packages('sams')

Monthly Downloads

189

Version

0.4.3

License

GPL-3

Maintainer

David B Dahl

Last Published

April 19th, 2022

Functions in sams (0.4.3)

Compute Probability Mass of a Partition Under the Two Parameter Chinese Restaurant Process (CRP)

Compute the Pochhammer Symbol (Rising Factorials) With Increment

simpleMergeSplit

Merge-Split Sampling for a Partition Using Uniformly Random Allocation

Independent Multivariate Normal Data (p = 18)

seqAllocatedMergeSplit_base

Base Functionality for the seqAllocatedMergeSplit Function

Correlated Multivariate Normal Data (p = 18)

Independent Multivariate Normal Data (p = 6)

partitionEntropy

Calculate the Entropy of a Set Partition

Compute the Posterior Pairwise Similarity for All Pairs of Items

Multivariate Independent Bernoulli Data (p = 18)

Conjugate Gibbs Sampler for a Partition

sizeOfLargestCluster

Calculate the Number of Items in the Largest Cluster of a Set Partition

Check if a Vector of Cluster Labels is in Canonical Form

Merge-Split Sampling for a Partition Based on Sequential Allocation Informed by Pairwise Similarities

Large Sample Multivariate Independent Bernoulli Data (p = 6)

psmMergeSplit_base

Base Functionality for the psmMergeSplit Function

joinExistingCluster

Join Item to an Existing Cluster

Count the Number of Clusters in a Set Partition

Multivariate Independent Bernoulli Data (p = 6)

transformedWeights

Enumerate Transformed Weights for Choosing i and j Non-Uniformly

seqAllocatedMergeSplit

Merge-split Sampling for a Partition Based on Sequential Allocation of Items

restrictedGibbsMergeSplit

Merge-Split Sampling for a Partition Based on Restricted Gibbs Scans

Coerce a Set Partition as Numeric Vectors of Cluster Labels to a List Structure

clusterProportions

Compute the Proportion of Items in Each Cluster for All Partitions

Plot Traces of Cluster Sizes

clusterWithItem

Identify Which Cluster Contains a Given Item

asClusterLabels

Coerce a Set Partition in List Structure to Numeric Vectors of Cluster Label

Coerce a Vector of Cluster Labels to Canonical Form

createNewCluster

Create a New Cluster with Given Item

Get theta Parameters from a Numeric Vector of Cluster Labels and Unique phi Values