SPECK (version 1.0.1)

ckmeansThreshold: Clustered thresholding of a vector.

Description

Performs thresholding for a vector of length \(m\) from a corresponding \(m x n\) reduced rank reconstructed (RRR) matrix. Thresholding of a vector is only performed if more than one cluster is identified using the Ckmeans.1d.dp::Ckmeans.1d.dp() function based on a one-dimensional dynamic programming clustering algorithm, which functions by minimizing the sum of squares of within-cluster distances from an element to its associated cluster mean. If more than one cluster is present, then the RRR output corresponding to the nonzero elements of the least-valued cluster, as identified by the cluster mean, is set to zero. All other values in the least and higher-valued clusters are retained.

Usage

ckmeansThreshold(rrr.vector, max.num.clusters = 4, seed.ckmeans = 2)

Value

• rrr.thresholded.vector - A thresholded vector of length \(m\).

• num.centers - Number of identified clusters.

• max.clust.prop - Proportion of samples with the specified maximum number of clusters.

Arguments

rrr.vector

Vector of length \(m\) from the corresponding \(m x n\) RRR matrix.

max.num.clusters

Maximum number of clusters for computation.

seed.ckmeans

Seed specified to ensure reproducibility of the clustered thresholding.

Examples

set.seed(10)
data.mat <- matrix(data = rbinom(n = 18400, size = 230, prob = 0.01), nrow = 80)
rrr.object <- randomizedRRR(counts.matrix = data.mat, rank.range.end = 60,
min.consec.diff = 0.01, rep.consec.diff = 2,
manual.rank = NULL, seed.rsvd = 1)
thresh.full.output <- ckmeansThreshold(rrr.vector = rrr.object$rrr.mat[,1],
max.num.clusters = 4, seed.ckmeans = 2)
head(thresh.full.output$rrr.thresholded.vector)
print(thresh.full.output$num.centers)
print(thresh.full.output$max.clust.prop)