coat | R Documentation |
Compositional-adjusted thresholding by doi.org/10.1080/01621459.2018.1442340 by Cao, Lin & Li (2018)
coat(
data,
lambda,
thresh = "soft",
adaptive = TRUE,
shrinkDiag = TRUE,
ret.icov = FALSE,
...
)
data |
a clr-transformed data or covariance matrix |
lambda |
threshold parameter(s) |
thresh |
"soft" or "hard" thresholding |
adaptive |
use adative-version of the lambda as in the original COAT paper. See details. |
shrinkDiag |
flag to exclude the covariance diagonal from the shrinkage operation |
ret.icov |
flag to also return the inverse covariance matrix (inverse of all thresholded COAT matrices) |
... |
Arguments to automatically calculating the lambda path. See details. |
If adaptive=TRUE
, and data
is a covariance matrix, the adaptive penalty is calculated by assuming the underlying data is jointly Gaussian in the infinite sample setting. The results may differ from the 'empirical' adaptive setting.
There are a few undocumented arguments useful for computing a lambda path on the fly:
lambda.max: Maximum lambda. Default: max absolute covariance
lambda.min.ratio: lambda.min is lambda.min.ratio*lambda.max is the smallest lambda evaluated. Defailt: 1e-3
nlambda: Number of values of lambda between lambda.max and lambda.min. Default: 30
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.