dot-getHurdle: Obtains logistic fits for presence/absence and fitted...

In HCBravoLab/Wrench: Wrench normalization for sparse count data

Description

This function is used to derive weights for feature-wise compositional estimates. Our (default) intention is to derive these based on average occurrences across the dataset, as just a function of sample depth, and not with particular relevance to groups.

Usage

.getHurdle(mat, hdesign = model.matrix(~-1 + log(colSums(mat))),
  pres.abs.mod = TRUE, thresh = FALSE, thresh.val = 1e-08, ...)

Arguments

mat	count matrix
hdesign	design matrix for the logistic; the default is usually sufficient.
pres.abs.mod	TRUE if glm regression is for presence or absence. FALSE if glm regression is for counts.
thresh	TRUE if numerically one/zero probability occurrences must be thresholded
thresh.val	if thresh is true, the numerically one/zero probability occurrences is thresholded to this value
...	other parameters

Value

A list with components:

• pi0.fit - list with feature-wise glm.fit objects

• pi0 - matrix with fitted probabilities

HCBravoLab/Wrench documentation built on May 13, 2019, 5:35 p.m.