loessByCol: Locally Weighted Mean By Column

In hiraksarkar/edgeR_fork: Empirical Analysis of Digital Gene Expression Data in R

Description

Smooth columns of matrix by non-robust loess curves of degree 0.

Usage

loessByCol(y, x=NULL, span=0.5)
locfitByCol(y, x=NULL, weights=1, span=0.5, degree=0)

Arguments

y	numeric matrix of response variables.
x	numeric covariate vector of length nrow(y), defaults to equally spaced.
span	width of the smoothing window, in terms of proportion of the data set. Larger values produce smoother curves.
weights	relative weights of each observation, one for each covariate value.
degree	degree of local polynomial fit

Details

Fits a loess curve with degree 0 to each column of the response matrix, using the same covariate vector for each column. The smoothed column values are tricube-weighted means of the original values.

locfitByCol uses the locfit.raw function of the locfit package.

Value

A list containing a numeric matrix with smoothed columns and a vector of leverages for each covariate value.

locfitByCol returns a numeric matrix.

Author(s)

Aaron Lun for loessByCol, replacing earlier R code by Davis McCarthy. Gordon Smyth for locfitByCol.

See Also

loess

Examples

y <- matrix(rnorm(100*3), nrow=100, ncol=3)
head(y)
out <- loessByCol(y)
head(out$fitted.values)

hiraksarkar/edgeR_fork documentation built on Dec. 20, 2021, 3:52 p.m.