locfitByCluster: Apply local regression smoothing to values within each...

In bumphunter: Bump Hunter

Description

Local regression smoothing with a gaussian kernal, is applied independently to each cluster of genomic locations. Locations within the same cluster are close together to warrant smoothing across neighbouring locations.

Usage

locfitByCluster(y, x = NULL, cluster, weights = NULL, minNum = 7,
               bpSpan = 1000, minInSpan = 0, verbose = TRUE)

Arguments

y	A vector or matrix of values to be smoothed. If a matrix, each column represents a sample.
x	The genomic location of the values in y
cluster	A vector indicating clusters of locations. A cluster is typically defined as a region that is small enough that it makes sense to smooth across neighbouring locations. Smoothing will only be applied within a cluster, not across locations from different clusters.
weights	weights used by the locfit smoother
minNum	Clusters with fewer than minNum locations will not be smoothed
bpSpan	The span used when locfit smoothing. (Expressed in base pairs.)
minInSpan	Only smooth the region if there are at least this many locations in the span.
verbose	Boolean. Should progress be reported?

Details

This function is typically called by smoother, which is in turn called by bumphunter.

Value

fitted	The smoothed data values
smoothed	A boolean vector indicating whether a given position was smoothed
smoother	always set to ‘locfit’.

Author(s)

Rafael A. Irizarry and Kasper D. Hansen

See Also

smoother, runmedByCluster, loessByCluster

Examples

dat <- dummyData()
smoothed <- locfitByCluster(y=dat$mat[,1], cluster=dat$cluster, bpSpan = 1000,
                         minNum=7, minInSpan=5)

bumphunter documentation built on Nov. 8, 2020, 4:59 p.m.