rMisbeta (version 1.0)

rMisbeta-package: A Robust Missing Imputation Method for Gene Expression and Metabolomics Data Analysis

Description

We developed a robust missing value imputation approach for gene expression and metabolomics data analysis using minimum beta-divergence method. This approach capable of handling both missing values and outliers, simultaneously.

Details

Package:	rMisbeta
Type:	Package
Version:	1.0
Date:	2020-10-03
License:	GPL (>=2.0)

Package rMisbeta has the six following functions:

Sim2Group():	This function generates the data from the one way ANOVA model for two groups.
OutMisDat():	This function returns the outliers and missing value incorporated data.
CalcMeanVar():	This function calculates the robust mean and variance from the the matrix in presence
	of outliers and missing values for function RobMeanVar().
RobMeanVar():	This function calculates the robust mean and variance from the the matrix in presence
	of outliers and missing values. The function RobMeanVar() also produces a
	weight called beta-weights for each of the values to detect the outliers in the dataset.
remat():	This function returns reconstructed data matrix by modifying the outliers and missing value
	using beta divergence method.
performance.eval():	This is the performance evaluation function. Which calculates TPR,TNR,FPR,PNR,AUC
	etc. as a measure of performance index.

Examples

# NOT RUN {
nG=1000
n1=n2=5
pde=0.1
Simdat=Sim2Group(nG,n1,n2,var0=0.1,pde=0.1) 
xx=Simdat$outmat
TrueDE=Simdat$DEtrue
MisOutdat<-OutMisDat(xx,pctOut=0.1,pctMis=0.1)
misdat_zero<-MisOutdat
misdat_zero[is.na(misdat_zero)]<-0
cl=rep(c(1,2),each=n1)

res=remat(MisOutdat,cl)
up_mat<-res$remat


pTtest_zero<-pTtest_beta<-NULL
for (j1 in 1:dim(xx)[1])
{
DataYY <- data.frame(YY =misdat_zero[j1,], FactorLevels = factor(cl))
DataYY2 <- data.frame(YY2=up_mat[j1,], FactorLevels2 = factor(cl))
pTtest_zero[j1] <-  t.test(YY~FactorLevels,data=DataYY, paired=FALSE)[[3]]
pTtest_beta[j1] <-  t.test(YY2~FactorLevels2,data=DataYY2, paired=FALSE)[[3]]
} 

TopDEn<-seq(nG*pde/10, pde*nG, length=10)

performance_zero<-performance.eval(pTtest_zero,TrueDE,TopDEn,decreasing=FALSE);
performance_beta<-performance.eval(pTtest_beta,TrueDE,TopDEn,decreasing=FALSE);
plot(performance_zero$FPR,performance_zero$TPR,type="o",
xlab="False Positive Rate",ylab="True Positive Rate",ylim=c(0,1))
points(performance_beta$FPR,performance_beta$TPR,type="o",col=2)
legend("bottomright", c('t_test_zero','t_test_rMisbeta'),lwd=1,cex=0.8,col=c(1,2))
# }