summaryDIRECT: Processing Posterior Estimates for Clustering Under DIRECT

In DIRECT: Bayesian Clustering of Multivariate Data Under the Dirichlet-Process Prior

Processing Posterior Estimates for Clustering Under DIRECT

Description

Function summaryDIRECT processes posterior estimates in the output files from DIRECT for clustering and parameter estimation.

Usage

summaryDIRECT(data.name, PERM.ADJUST = FALSE)

Arguments

data.name	A character string indicating the prefix of output files.
PERM.ADJUST	If TRUE, add 1 to labels of mixture components such that the labels start from 1 instead of 0.

Details

Output files from DIRECT include MCMC samples before relabeling and permuted labels of mixture components after relabeling. Function summaryDIRECT uses permuted labels stored in output file *_mcmc_perms.out to reorganize the MCMC samples stored in other output files *_mcmc_cs.out, *_mcmc_pars.out and *_mcmc_probs.out. It defines each mixture component as a cluster.

Value

A list with components:

nitem	The number of items in the data.
nclust	The number of inferred clusters.
top.clust.alloc	A vector of length nitem, each component being the maximum posterior probability of allocating the corresponding item to a cluster.
cluster.sizes	Vector of cluster sizes.
top.clust.labels	An integer vector of labels of inferred clusters. The integers are not necessarily consecutive; that is, an inferred mixture component that is associated with items at small posterior allocation probabilities is dropped from the final list of cluster labels.
top2allocations	A data frame containing "first", the most likely allocation; "second", the second most likely allocation; "prob1", the posterior allocation probability associated with "first"; and "prob2", the posterior allocation probability associated with "second".
post.alloc.probs	A nitem-by-nclust matrix of mean posterior allocation probability matrix.
post.clust.pars.mean	A matrix of nclust rows. Each row, corresponding to an inferred cluster, contains the posterior mean estimates of cluster-specific parameters.
post.clust.pars.median	A matrix of nclust rows. Each row, corresponding to an inferred cluster, contains the posterior median estimates of cluster-specific parameters.
misc	A list containing two components: post.pars.mean: Matrix each row of which contains the posterior mean estimates of parameters for a mixture component. post.pars.median: Matrix each row of which contains the posterior median estimates of parameters for a mixture component.

Author(s)

Audrey Q. Fu

References

Fu, A. Q., Russell, S., Bray, S. and Tavare, S. (2013) Bayesian clustering of replicated time-course gene expression data with weak signals. The Annals of Applied Statistics. 7(3) 1334-1361.

See Also

DIRECT for what output files are produced.

simuDataREM for simulating data under the mixture random-effects model.

Examples

## See example in DIRECT.

DIRECT documentation built on Sept. 8, 2023, 5:45 p.m.