plotBIC {Ckmeans.1d.dp}

Plot Bayesian Information Criterion as a Function of Number of Clusters

Description

Plot Bayesian information criterion (BIC) as a function of the number of clusters obtained from optimal univariate clustering results returned from Ckmeans.1d.dp. The BIC normalized by sample size (BIC/n) is shown.

Usage

plotBIC(
  ck, xlab="Number of clusters k",
  ylab = "BIC/n", type="b",
  sub=paste("n =", length(ck$cluster)),
  main=paste("Bayesian information criterion",
             "(normalized by sample size)", sep="\n"),
  ...
)

Arguments

ck

an object of class Ckmeans.1d.dp returned by Ckmeans.1d.dp.

xlab

a character string. The x-axis label for the plot.

ylab

a character string. The x-axis label for the plot.

type

the type of plot to be drawn. See plot.

main

a character string. The title for the plot.

sub

a character string. The subtitle for the plot.

...

arguments passed to plot function in package graphics.

Details

The function visualizes the input data as sticks whose heights are the weights. It uses different colors to indicate optimal k-means clusters. The method to calcualte BIC based on Gaussian mixture models estimated on a univariate clustering is described in (Song and Zhong 2020).

Value

An object of class "Ckmeans.1d.dp" defined in Ckmeans.1d.dp.

Author(s)

Joe Song

References

Song M, Zhong H (2020). “Efficient weighted univariate clustering maps outstanding dysregulated genomic zones in human cancers.” Bioinformatics, 36(20), 5027–5036. doi:10.1093/bioinformatics/btaa613.

Examples

# Example: clustering data generated from a Gaussian mixture
#          model of two components
x <- rnorm(50, mean=-1, sd=0.3)
x <- append(x, rnorm(50, mean=1, sd=0.3) )
res <- Ckmeans.1d.dp(x)
plotBIC(res)

y <- (rnorm(length(x)))^2
res <- Ckmeans.1d.dp(x, y=y)
plotBIC(res)