flowMeans (version 1.32.0)

flowMeans: flowMeans

Description

Finds a good fit to the data using k-means clustering algorithm. Then merges the adjacent dense spherical clusters to find non-spherical clusters.

Usage

flowMeans(x, varNames=NULL, MaxN = NA, NumC = NA, iter.max = 50, nstart = 10,
Mahalanobis = TRUE, Standardize = TRUE, Update = "Mahalanobis", OrthagonalResiduals=TRUE,
MaxCovN=NA, MaxKernN=NA, addNoise=TRUE)

Arguments

x

A matrix, data frame of observations, or object of class flowFrame. Rows correspond to observations and columns correspond to variables.

varNames

A character vector specifying the variables (columns) to be included in clustering. When it is left unspecified, all the variables will be used.

MaxN

Maximum number of clusters. If set to NA (default) the value will be estimated automatically.

NumC

Number of clusters. If set to NA (default) the value will be estimated automatically.

iter.max

The maximum number of iterations allowed.

nstart

The number of random sets used for initialization.

Mahalanobis

Boolean value. If TRUE (default) mahalanobis distance will be used. Otherwised, euclidean distance will be used.

Standardize

Boolean value. If TRUE (default) the data will be transformed to the [0,1] interval.

Update

String value. If set to "Mahalanobis" the distance function will be updated at each merging iteration with recalculating mahalanobis distances. If set to "Mean" the distance matrix will be updated after each merging step with averaging. If set to "None" the distance matrix will not be updated.

MaxCovN

Maximum number of points, used for calculating the covariance. If set to NA (default), all the points will be used.)

MaxKernN

Maximum number of points, used for counting the modes using kernel density estimation. If set to NA (default), all the points will be used.)

addNoise

Boolean value. Determines if uniform noise must be added to the data to prevent singularity issues or not.

OrthagonalResiduals

Boolean value, indicates if the residuals must be transformed to orthagonal distance or not.

Value

Label

A vector of integers indicating the cluster to which each point is allocated.

Labels

A list of vectors of integers indicating the cluster to which each point is allocated at each merging iteration.

Mats

A list of distance matrixes between clusters at every merging iteration.

MaxN

Maximum number of clusters

Mins

A vector of integers indicating the distance between the two clusters chosen to be merged at every iteration.

MinIndex

Index of the merging step that produced the final results.

Line1

First regression line used for finding the changepoint for stopping the merging process.

Line2

Second regression line used for finding the changepoint for stopping the merging process.

Details

If Mahalanobis distance is not used (i.e., Mahalanobis=FALSE) then the Update value cannot be set to Mahalanobis (i.e., Update="Mahalanobis")

Examples

library(flowMeans)
data(x)
res <- flowMeans(x, c("FL1.H", "FL2.H", "FL3.H", "FL4.H"), MaxN=10)
plot(x[,c(3,4)], res, c("FL1.H", "FL2.H"))