README.md

In HGray384/TAS: Target-Averaged Linear Shrinkage (TAS)

TAS

Target-Averaged linear Shrinkage R-package (dev version)

Implementation of TAS from Gray et al. (2018) for high-dimensional covariance matrix estimation. Conceptually, TAS performs linear shrinkage by weighting multiple target matrices by their log-marginal likelihood values in a Gaussian conjugate (inverse Wishart prior) model. This is a data-driven way of either performing regularisation through a restricted target matrix, or using a strongly informative prior through a data-derived target matrix for potentially greater estimation accuracy. The model weighting procedure eliminates the need to choose a single target matrix, which is the case in standard linear shrinkage.

See documentation files for information on the usage of TAS.

This version can be directly installed from the R console using:

# install.packages("devtools")
devtools::install_github("HGray384/TAS")

HGray384/TAS documentation built on Dec. 14, 2020, 8:41 p.m.