makeRankFull: Enforce full rank design matrix

In nbenn/singleCellFeatures: Analyze single cell features stored in an openBIS instance

Description

Given a design matrix (if the response is also included it has to be named "Response"), rank deficiency is corrected by dropping zero variance variables and one of each pair of highly correlated variables to make the design matrix invertible.

Usage

makeRankFull(data, verbose = FALSE)

Arguments

data	Matrix/data frame holding the design matrix and optionally a response vector named "Response"

Value

A subset of the original data structure (some of the variables are removed)

Examples

# get gene locations
mtor.loc <- findWells(experiments="brucella-du-k1", contents="MTOR")
scra.loc <- findWells(plates=sapply(mtor.loc, getBarcode),
                      contents="SCRAMBLED", well.names="G23")
# combine for faster fetching
data     <- getSingleCellData(list(mtor.loc[[1]], scra.loc[[1]]))
mtor.dat <- meltData(cleanData(data[[1]]$H6))
scra.dat <- meltData(cleanData(data[[1]]$G23))
# prepare data for glm
data <- prepareDataforGlm(mtor.dat$mat$Cells, scra.dat$mat$Cells)
data <- makeRankFull(data)
# run glm (enforcing full rank might be advisbale)
model <- glm("Response ~ .", binomial, data$train)

nbenn/singleCellFeatures documentation built on May 23, 2019, 12:24 p.m.