fitZifNetwork: Fit Meinhousen-Buhlmann to a SingleCellAssay object

In amcdavid/SingleCellAnalysis: Tools for Analysis of Zero-Inflated, Continuous Data

Description

Regresses the dichotomous and continuous components of each gene in sc on every other gene. additive.effects from cData(sc) are included unpenalized. Currently only gene.predictors as 'zero.inflated' is supported. ... is passed along to cv.glmnet, see documentation there.

Usage

fitZifNetwork(sc, additive.effects, min.freq = 0.05,
  gene.predictors = "zero.inflated", precenter = TRUE,
  precenter.fun = scale, response = "hurdle", modelSelector,
  onlyReturnFitter = FALSE, ...)

Arguments

sc	SingleCellAssay object on a thresholded layer
additive.effects	character vector, possibly using formula syntax of columns from cData(sc) to be included as unpenalized terms.
min.freq	genes below this frequency are excluded as predictors and dependent variables
gene.predictors	ignored
precenter	How should centering/scaling be done with respect to continuous regressions. TRUE if centering should be done with respect to all cells; FALSE if centering should be done only with respect to expressed cells When precenter=TRUE, cv.glmnet will not standardize.
precenter.fun	a function called to center the expression matrix prior to calling glmnet
response	a character vector, one of 'zero.inflated', 'hurdle', or 'cg.regression'
modelSelector	a function called gene and component-wise on each fit, that should return an index to the glmnet lambda sequence for that gene and component
onlyReturnFitter	if TRUE, return an undocumented fitter function that is internally called on each gene/component.
...	passed to cv.glmnet

Value

2-D list of cv.glmnet objects with attributes

amcdavid/SingleCellAnalysis documentation built on May 10, 2019, 10:27 a.m.