Robust adaptive lasso based on robust correlation estimates

Usage

covlasso(
  x,
  y,
  cor.method = "gaussrank",
  scale.method = "qn",
  center.method = "median",
  adaptive = TRUE
)

Arguments

x: input design matrix
y: input response vector
cor.method: could be "pearson" or "gaussrank"
scale.method: "sd" or "qn"
adaptive: adaptive regularization penalties
std: If TRUE the robust correlation matrix is used, if FALSE the robust covariance matrix is used.
cormatrix: you could also use a correlation matrix as input
scale: put in scales if you use cormatrix

Value

betahat_opt, the optimal estimation of beta obtained from this algorithm

lambda_opt is the optimal tuning parameter and sigma_opt is the optimal estimation of sigma.

The output also includes the estimated correlation matrix, the estimated covariance matrix and et cetera from the covf function.

Examples

dat = genevar()
y = dat$y
x = dat$x
fit = covlasso(x,y)
fit$betahat_opt
#>    const.3      beta1      beta2      beta3      beta4      beta5      beta6 
#> -0.6848624  0.0000000  2.4248406  0.5241094  2.2935756  0.0000000  0.0000000 
#>      beta7      beta8      beta9     beta10     beta11     beta12     beta13 
#>  0.0000000  0.0000000  0.0000000  0.0000000  0.0000000  0.0000000  0.0000000 
#>     beta14     beta15     beta16     beta17     beta18     beta19     beta20 
#>  0.0000000  0.0000000  0.0000000  0.0000000  0.0000000  0.0000000  0.0000000