CORREP (version 1.38.0)

cor.unbalance: Multivariate Correlation Estimator (Unequal Number of Replicates)

Description

cor.unbalance estimates correlation from replicated data of unequal number of replicates. different from cor.balance, cor.unbalance takes a pair of variables at a time because of unequal number of replicates. the variance of each row of the data MUST equal to 1 (see example below)

Usage

cor.unbalance(x, m1, m2)

Arguments

x

data matrix, column represents samples (conditions), and row represents variables (genes), see example below for format information

m1

number of replicates for one variable (gene)

m2

number of replicates for another variable (gene)

Value

A correlation matrix containing only one distinct correlation coefficient for the pair of variables (genes)

Details

The multivariate correlation estimator assumes replicated omics data are iid samples from the multivariate normal distribution. It is derived by maximizing the likelihood function. Note that the off-diagonal elements in the returned correlation matrix (G by G) is the average of off-diagonals of MLE of correlation matrix of a pair of variables (m1+m2 by m1+m2).

References

Zhu, D and Li Y. 2007. Multivariate Correlation Estimator for Inferring Functional Relationships from Replicated 'OMICS' data. Submitted.

See Also

cor.unbalance, cor

Examples

library("CORREP")
d0 <- NULL
for(l in 1:10)
d0 <- rbind(d0, rnorm(8))
## The simulated data corresponds to the real-world data of 2 genes and 10 conditions, gene expression
## profiles were replicated 3 and 5 times. 
## Note this function can only take calculate correlation matrix between two genes at a time.
d0<- t(d0)
## This step is to make the standard deviation of each replicate equal to 1
## so that we can model the covariance matrix as correlation matrix.
d0.std <- apply(d0, 1, function(x) x/sd(x))
M <- cor.unbalance(t(d0.std), m1=3, m2=5)