genSimData.BayesNormal: Generating simulated data set from conditional normal...

In iCheck: QC Pipeline and Data Analysis Tools for High-Dimensional Illumina mRNA Expression Data

Description

Generating simulated data set from conditional normal distributions.

Usage

genSimData.BayesNormal(
  nCpGs, 
  nCases, 
  nControls,
  mu.n = -2,
  mu.c = 2,
  d0 = 20, 
  s02 = 0.64,
  s02.c = 1.5,
  testPara = "var", 
  outlierFlag = FALSE,
  eps = 0.001, 
  applier = lapply)

Arguments

nCpGs	integer. Number of genes.
nCases	integer. Number of cases.
nControls	integer. Number of controls.
mu.n	numeric. mean of the conditional normal distribution for controls. See details.
mu.c	numeric. mean of the conditional normal distribution for cases. See details.
d0	integer. degree of freedom for scale-inverse chi squared distribution. See details.
s02	numeric. scaling parameter for scale-inverse chi squared distribution for controls. See details.
s02.c	numeric. scaling parameter for scale-inverse chi squared distribution for cases. See details.
testPara	character string. indicating if the test is for testing equal mean, equal variance, or both.
outlierFlag	logical. indicating if outliers would be generated. If outlierFlag=TRUE, then we followed Phipson and Oshlack's (2014) simulation studies to generate one outlier for each CpG site by replacing the DNA methylation level of one diseased subject by the maximum of the DNA methylation levels of all CpG sites.
eps	numeric. if |mean0-mean1|<eps then we regard mean0=mean1. Similarly, if |var0-var1|<eps then we regard var0=var1. mean0 and var0 are the mean and variance of the chi squared distribution for controls. mean1 and var1 are the mean and variance of the chi squared distribution for cases.
applier	function name to do apply operation.

Details

Based on Phipson and Oshlack's (2014) simulation algorithm. For each CpG site, variance of the DNA methylation was first sampled from an scaled inverse chi-squared distribution with degree of freedom d_0 and scaling parameter s_0^2: σ^2_i ~ scale-inv χ^2(d_0, s_0^2). M value for each CpG was then sampled from a normal distribution with mean μ_n and variance equal to the simulated variance σ^2_i. For cases, the variance was first generated from σ^2_{i,c} ~ scale-inv χ^2(d_0, s_{0,c}^2). M value for each CpG was then sampled from a normal distribution with mean μ_c and variance equal to the simulated variance σ^2_{i,c}.

Value

An ExpressionSet object. The phenotype data of the ExpressionSet object contains 2 columns: arrayID (array id) and memSubj (subject membership, i.e., case (memSubj=1) or control (memSubj=0)). The feature data of the ExpressionSet object contains 4 elements: probe (probe id), gene (psuedo gene symbol), chr (psuedo chromosome number), and memGenes (indicating if a gene is differentially expressed (when testPara="mean") or indicating if a gene is differentially variable (when testPara="var") ).

Author(s)

Weiliang Qiu <[e-mail]>, Brandon Guo <[e-mail]>, Christopher Anderson <[e-mail]>, Barbara Klanderman <[e-mail]>, Vincent Carey <[e-mail]>, Benjamin Raby <[e-mail]>

References

Phipson B, Oshlack A. DiffVar: A new method for detecting differential variability with application to methylation in cancer and aging. Genome Biol 2014; 15:465

Examples

    # generate simulated data set from conditional normal distribution
    set.seed(1234567)
    es.sim = genSimData.BayesNormal(nCpGs = 100, 
      nCases = 20, nControls = 20,
      mu.n = -2, mu.c = 2,
      d0 = 20, s02 = 0.64, s02.c = 1.5, testPara = "var",
      outlierFlag = FALSE, 
      eps = 1.0e-3, applier = lapply) 
    print(es.sim)

iCheck documentation built on Nov. 8, 2020, 11:09 p.m.