lambda.TargetD: Shrinkage Estimation of a Covariance Matrix Toward an...

In GGMridge: Gaussian Graphical Models Using Ridge Penalty Followed by Thresholding and Reestimation

Shrinkage Estimation of a Covariance Matrix Toward an Identity Matrix

Description

Estimation of a weighted average of a sample covariance (correlation) matrix and an identity matrix.

Usage

lambda.TargetD(x)

Arguments

x	Centered data for covariance shrinkage and standardized data for correlation shrinkage.

Details

An analytical approach to the estimate ridge parameter.

Value

The estimates of shrinkage intensity.

Author(s)

Min Jin Ha

References

Schafer, J. and Strimmer, K. (2005). A shrinkage approach to large-scale covariance matrix estimation and implications for functional genomics. Statistical Applications in Genetics and Molecular Biology, 4, 32.

Ha, M. J. and Sun, W. (2014). Partial correlation matrix estimation using ridge penalty followed by thresholding and re-estimation. Biometrics, 70, 762–770.

Examples

###############################
 # Simulate data
 ###############################
 simulation <- simulateData(G = 100, etaA = 0.02, n = 50, r = 10)
 dat <- simulation$data[[1L]]
 stddat <- scale(x = dat, center = TRUE, scale = TRUE)
 
 shrinkage.lambda <- lambda.TargetD(x = stddat)
 
 ###############################
 # the ridge parameter
 ###############################
 ridge.lambda <- shrinkage.lambda / (1.0 - shrinkage.lambda)
 
 ###############################
 # partial correlation matrix
 ###############################
 partial <- solve(cor(dat) + ridge.lambda * diag(ncol(dat)))
 partial <- -scaledMat(x = partial)
 

GGMridge documentation built on Nov. 25, 2023, 1:08 a.m.