CIlambda: Confidence intervals for lambda

In isglobal-brge/lambda: Computes the agreement measure between two experiments across numerous conditions

Description

Computes the confidence intervals of lambda from bootstrapping

Usage

CIlambda(E1, E2, alpha, nsam=1000, mc.cores=1, benchmark = FALSE, verbose=FALSE)

Arguments

E1	Matrix of k conditions (tissues) and N observations (gene-pair correlations) for rater 1 (experiment 1)
E2	Matrix of k conditions (tissues) and N observations (gene-pair correlations) for rater 2 (experiment 2)
alpha	confidence level of CI
nsam	number of resamplings with replacements
mc.cores	number cores for parallelization
benchmark	a logical (default FALSE) that computes the agreement measure for benchmarking experiment 2 (columns) against experiment 1 (rows).
verbose	logical. If TRUE it prints the sampling number during computation.

Details

The function computes the CI of lambda at alpha level from re-sampling with replacement the rows of E1 and E2. E1 and E2 correspond to the observations made by each rater across multiple conditions (columns). Each observation is assumed to be the correlation between two variables.

For gene networks the variables may be the correlation of expression levels between two genes across subjects. In this case, the first column in E1 contains, for instance, all non-redundant pair-wise correlations among the genes in the network for condition (tissue) 1 in experiment 1.

Value

CI	confidence intervals

Author(s)

A Caceres

References

Caceres, A and Gonzalez JR, A measure of agreement across numerous conditions: Assessing when changes in network connectivities across tissues are functional

Examples

#simulate E1 and E2 from a multivariate distribution
library(MASS)
library(parallel)

nlevels<-5 #total conditions
ntrue<-3 #number of reliable conditions
cortrue<-0.5 #correlations of reliable conditions
basecor<-0.1 #background correlation

ngenes<-10 #number of genes in network
nonredundantcor<-ngenes*(ngenes-1)/2 #number of non-redundant correlations

#specify covariance matrix for simulation
covmat<-matrix(basecor, ncol=nlevels*2, nrow=nlevels*2)
  
for(ind in 1:nlevels)
  {  
    covmat[(2*ind-1):(2*ind), (2*ind-1):(2*ind)] <- matrix(c(1,cortrue,cortrue,1), ncol = 2)
    if(ind > ntrue)
      covmat[(2*ind-1):(2*ind), (2*ind-1):(2*ind)] <- matrix(c(1,basecor,basecor,1), ncol = 2)
  }
  
simul <- mvrnorm(nonredundantcor, mu = rep(0,nlevels*2), Sigma = covmat, empirical = TRUE)
### 
  
#45 non-redundant gene correlation pairs from a 10 gene-network on 5 conditions, 3 of which are inter experiment reliable (3/5=0.6)

E1 <- simul[,seq(1,ncol(simul),2)] #experiment 1
E2 <- simul[,seq(1,ncol(simul),2)+1] #experiment 2
  
CIlambda(E1=E1, E2=E2, alpha=0.05, nsam=100, mc.cores=2)
    

isglobal-brge/lambda documentation built on May 26, 2019, 12:31 p.m.