Robocov_cor_slack: Robocov correlation estimation using box constraints on...

In kkdey/Robocov: Robust Estimation of covariance matrix and conditional graphs

Description

A robust estimation of correlation matrix for data with missing entries using box constraint on the difference between the population correlation matrix and pairwise sample correlation matrix, with added L-1 penalty on the slack variables. This is a more flexible version of the Robocov_cor function.

Usage

Robocov_cor_slack(data_with_missing, alpha = 1, loss = c("lasso",
  "ridge", "elasticnet"))

Arguments

data_with_missing	Samples by features data matrix. May contain missing entries (NA) values.
alpha	The tuning parameter for the L-1 shrinkage of the slack variables.
loss	Specify if we minimize L-1 ('lasso'), L-2 ('ridge') or elastic-net ('elasticnet') loss functions.

Examples

data("sample_by_feature_data")
out = Robocov_cor_slack(sample_by_feature_data, alpha = 1)
corrplot::corrplot(as.matrix(out), diag = FALSE,
        col = colorRampPalette(c("blue", "white", "red"))(200),
        tl.pos = "td", tl.cex = 0.4, tl.col = "black",
        rect.col = "white",na.label.col = "white",
        method = "color", type = "upper")

kkdey/Robocov documentation built on June 12, 2020, 11:34 a.m.