How to install

The development version from GitHub:

# install.packages("remotes")
remotes::install_github("AlexisDerumigny/UniversalShrink")

1. Functions for estimation of the covariance matrix

cov_analytical_NL_shrinkage() and cov_quadratic_inverse_shrinkage(): perform estimation of the covariance matrix using non-linear shrinkage. Both estimators are optimal for the Frobenius norm (asymptotically).
cov_higher_order_shrinkage(): perform higher-order shrinkage of the covariance matrix.

Moore_Penrose(): Moore-Penrose estimator of the precision matrix, obtained by computing the Moore-Penrose inverse of the sample covariance matrix.
Moore_Penrose_shrinkage(): perform a first-order shrinkage of the Moore-Penrose estimator of the precision matrix, towards an arbitrary (fixed) target such as the identity matrix.
Moore_Penrose_higher_order_shrinkage(): estimate the precision matrix via a polynomial in the Moore-Penrose estimator of the precision matrix.

ridge(): the usual ridge estimator
ridge_shrinkage(): perform first-order shrinkage of the Ridge estimator towards an arbitrary (fixed) target such as the identity matrix.
ridge_higher_order_shrinkage(): perform higher-order shrinkage of the Ridge estimator
ridge_shrinkage_Wang()

From a given precision matrix, one can get the optimal portfolio weights (GMV for Global Minimum Variance) using the function GMV_PlugIn(). This can also be used by giving as input an estimator of the precision matrix, which then will give as output an estimator of the optimal portfolio weights.
GMV_Moore_Penrose(): using the Moore-Penrose inverse Moore_Penrose() of the precision matrix as an input for the plug-in estimation.
GMV_Moore_Penrose_shrinkage(): performs a first-order shrinkage of the estimator given by GMV_Moore_Penrose() towards an arbitrary (fixed) portfolio such as the equally weighted portfolio.