estimateBPMatrix: Estimate parameters of beta-Poisson models for a data matrix
In nghiavtr/BPSC: Beta-Poisson Model for Single-Cell RNA-seq Data Analyses

Description Usage Arguments Value Examples

Estimate parameters of beta-Poisson models for a data matrix

1
2
3

estimateBPMatrix(dataMat, para.num = 4, tbreak.num = 10, fout = NULL,
  break.thres = 10, estIntPar = TRUE, extreme.quant = NULL,
  useExt = FALSE, min.exp = 1e-04, useParallel = FALSE)

`dataMat`	Data matrix that needs to be modelled
`para.num`	Mode of beta-Poisson model: 3, 4 (default) or 5 parameters
`tbreak.num`	Number of breaks for binning
`fout`	A *.RData file name to export results
`break.thres`	A parameter setting of `getTbreak` function
`estIntPar`	An option to allow estimating initial parameters for the model from only expressed values
`extreme.quant`	A quantile probability to remove extrem values (outliers) higher than the quantile. If extreme.quant=NULL, no elimination of outliers is done
`useExt`	A parameter setting of `getTbreak` function that allows to extend the last bin to infinity or not
`min.exp`	A threshold for minimum expressed values. If a expression is less than min.exp, it is set to be zero
`useParallel`	An option to allow using parallel computing

A list of optimal models corresponding to the rows of the matrix. Each model consists of optimal parameters, X2 test results (X2 and PVAL), etc..

set.seed(2015)
#create random data matrix from a beta-poisson model
N=10
alp=sample(100,N,replace=TRUE)*0.1;
bet=sample(100,N,replace=TRUE)*0.1;
lam1=sample(100,N,replace=TRUE)*10;
lam2=sample(100,N,replace=TRUE)*0.01;
n=100
bp.mat=NULL
for (i in 1:N)
  bp.mat=rbind(bp.mat,rBP(n,alp=alp[i],bet=bet[i],lam1=lam1[i],lam2=lam2[i]))
#Estimate parameters from the data set
mat.res=estimateBPMatrix(bp.mat,para.num=4,fout=NULL,estIntPar=FALSE,useParallel=FALSE)
#In this function, user can also set estIntPar=TRUE to have better estimated beta-Poisson
#models for the generalized linear model. However, a longer computational time is required.