landmarkMatrix: Compute and cluster high density regions in 1D

Description

This functions first identifies high-density regions for each flowFrame in a flowSet and subsequently tries to cluster these regions, yielding the landmarks matrix that needs to be supplied to landmarkreg. The function is considered to be internal.

Usage

landmarkMatrix(data, fres, parm, border=0.05, peakNr=NULL, densities =
NULL, n = 201, indices=FALSE)

Arguments

data

A flowSet.

fres

A list of filterResultList objects generated by a filtering opration using a curv1Filter. Each list item represents the results for one of the flow parameters in parm.

parm

Character scalar of flow paramater to compute landmarks for.

border

A numeric in [0,1]. Ignore all high-density regions with mean values in the extreme percentiles of the data range.

peakNr

Force a fixed number of peaks.

densities

An optional matrix of y values of the density estimates for the flowSet. If this is not present, density estimates will be calculated by the function.

Number of bins used for the density estimation.

indices

Return matrix of population indices instead of landmark locations. These indices can be used to point into the populations identified by the curv1Filter.

Value

A matrix of landmark locations. Columns are landmarks and rows are flowFrames.

Details

In order to normalize the data using the landmarkreg function in the fda, a set of landmarks has to be computed for each flowFrame in a flowSet. The number of lansmarks has to be the same for each frame. This function identifies high-density regions in each frame, computes a simple clustering and returns a matrix of landmark locations. Missing landmarks of individual frames are substituted by the mean landmark location of the respective cluster.

Examples

Run this code


data(GvHD)
tmp <- list("FSC-H"=filter(GvHD[1:3], curv1Filter("FSC-H")))
res <-  flowStats:::landmarkMatrix(GvHD[1:3], tmp, "FSC-H")