sigma_estim_tlasso: t-Lasso (TLASSO) Covariance Estimation

In antshi/CovEstim: (High-dimensional) Covariance Estimation

Description

Computes the TLASSO estimator of the covariance matrix.

Usage

sigma_estim_tlasso(
  data,
  rho = NULL,
  pendiag_log = FALSE,
  df = 3,
  tol = 1e-05,
  maxit = 10000,
  symmetric_log = TRUE,
  theta_init = NULL
)

Arguments

data	an nxp data matrix.
rho	a double, the non-negative regularization parameter ρ for lasso. ρ=0 means no regularization. Can be a scalar (usual) or a symmetric p by p matrix, or a vector of length p. In the latter case, the penalty matrix has jkth element √{ρ_j*ρ_k}. Default value is 0.
pendiag_log	a logical, indicating whether the diagonal of the sample covariance matrix is to be penalized (TRUE) or not (FALSE). Default value is FALSE.
df	an integer, indicating the degrees of freedom of the assumed t-distribution. Default value is 3.
tol	a double, indicating the tolerance for the glasso algorithm. Default value is set to 1e-05.
maxit	an integer, indicating the maximum number of iterations for the glasso algorithm. Default value is set to 10000.
symmetric_log	a logical, indicating whether the output should be a symmetric matrix (TRUE) or not necessarily (FALSE). Default value is set to TRUE.
theta_init	a pxp initial matrix for the inverse of the covariance matrix. Default value is NULL and the sample inverse for the t-distribution is used.

Details

The TLASSO estimator is elaborated in detail in \insertCitefinegold2011robust;textualCovEstim. Originally developed by \insertCitetorri2019sparse;textualCovEstim.

Value

a list with the following entries

• a pxp estimated covariance matrix.

• an estimation specific tuning parameter, here the lasso penalty.

References

\insertAllCited

Examples

data(sp200)
sp_rets <- sp200[,-1]
sigma_tlasso <- sigma_estim_tlasso(sp_rets, rho=0.001)[[1]]

antshi/CovEstim documentation built on Nov. 13, 2020, 2:25 p.m.