NBID: negative binomial models allowing independent dispersions

In chenlab-sj/nbid: Differential expression analysis using negative binomial with independent dispersions (NBID)

Description

This function tests the following hypothesis H0: the same mu for different groups H1: different mu for different groups In both cases, the dispersion is group specific. The dispersion is estimated only for the full model and then used for the reduced model

Usage

NBID(count, groups, countPerCell, covariates, sizeFactor = NULL,
  singleDispersion = F, dispMethod = "poisson-ML")

Arguments

count	the vector of gene counts
groups	the vector of group information
covariates	the covariaites
sizeFactor	the normalization factor, the effective count used will be the countPerCell * sizeFactor. The size factor from package scran needs to be divided by countPerCell first, i.e., the size factor from scran is used as the effective count directly.
singleDispersion	whether to use one single dispersion for all cells. All groups will share this single dispersion parameter, i.e., no independent dispersions are used
dispMethod	the method to estimate the dispersion
countPercell	the total UMI of each cell

Value

a vector of the following:

LR: likelihood rato

beta: the effect parameter

dispersionOutput: dispersions of different groups

chenlab-sj/nbid documentation built on Nov. 4, 2019, 8:50 a.m.