DESCEND.control: The control parameters of the function 'runDescend' and...

In jingshuw/descend: Gene Expression Distribution Deconvolution for Single Cell RNA-seq

Description

The control parameters of the function runDescend and deconvSingle

Usage

DESCEND.control(n.points = 50, nStart = 2, nStart.lrt = 2,
  max.sparse = c(0.99, 20), LRT.Z.select = T, LRT.Z0.select = T,
  LRT.Z.values = 0, zeroInflate = T, dense.add.0 = T,
  only.integer = F, discrete.quantile.max = 0.98,
  rel.info.range = c(5e-04, 0.01), rel.info.guide = 0.005,
  c0.start = 1, aStart = 1, bStart = 0, gStart = 0,
  start.sd = 0.2, penalty.Z0 = T, pDegree = 5, max.c0.iter = 5,
  c0.min = 1e-05)

Arguments

n.points	number of discritized points of the underlying true expression distribution. Default is 50
nStart	number of random starts for the optimization problem (as it is non-convex) to find the global minimum. Default is 2
nStart.lrt	number of random starts for the unpenalized optimization problem for likelihood ratio testing
max.sparse	a vector of 2 indicating the maximum sparsity allowed for a gene to be computed. The first element is the fraction of zero-counts allowed, the second element is the minimum number of non-zero counts. Both criteria should be satisfied. Default is (0.99, 20). For studying nonzero fraction, one should increase the threshold to get estimates with acceptable accuracy.
LRT.Z.select	a vector of length 1 or the number of columns of Z indicating for which column of Z the coefficients are tested against the corresponding value in LRT.Z.values using LRT. Default is TRUE, meaning that all columns are tested when LRT tests are performed.
LRT.Z0.select	a vector of length 1 or the number of columns of Z0 indicating for which column of Z0 the coefficients are tested against 0 using LRT. Default is TRUE, meaning that all columns are tested when LRT tests are performed.
LRT.Z.values	a vector of length 1 or the number of columns of Z showing the values that LRT tests on coefficients of Z are tested against. Default value is 0, meanings that all tests are tested against 0.
zeroInflate	whether to include the zero inflation part to the deconvolved distribution. Default is TRUE
dense.add.0	whether smooth the density at 0 into density.points in the output DESCEND object
only.integer	whether set the discrete points to be integers. Default is FALSE
discrete.quantile.max	The maximum quantile of the observed distribution used to build discrete points for the deconvolved distribution
rel.info.range	the relative information range allowed for finding the optimal tuning parameter c0
rel.info.guide	one parameter inside the rel.info.range controling for the searching process of c0
c0.start	the starting value of c0. Default is 1
aStart	the starting values of the spline coefficients of the deconvolved distribution
bStart	the starting values of the coefficients of Z0
gStart	the starting values of the coefficients of Z
start.sd	standard deviation of the random starting values when nStart > 1
penalty.Z0	whether add penalty to the coefficients of Z0 or not in optimization. Default is TRUE
pDegree	degree of the spline bases. Default is 5
max.c0.iter	maximum iteration allowed to find the optimal c0
c0.min	minimum c0 allowed to avoid degeneration

jingshuw/descend documentation built on Nov. 2, 2021, 4:23 p.m.