Robocov_lowrank: Robocov correlation matrix estimation of data with low rank +...
In kkdey/Robocov: Robust Estimation of covariance matrix and conditional graphs

Description Usage Arguments Examples

View source: R/Robocov_lowrank.R

Performs a nuclear norm penalization of the correlation matrix with low-rank + noise structure when the data has missing entries.

1 2	Robocov_lowrank(data_with_missing, alpha = 0.01, stepsize = 1, max_iter = 1000, tol = 1e-04, verbose = TRUE)

`data_with_missing`	The samples by features data matrix. May contain NA values.
`alpha`	The tuning parameter for the gradient descent(GD) iteration update.
`stepsize`	The stepsize for the gradient descent algorithms.
`max_iter`	The maximum number of iterations for the GD.
`tol`	The tolerance level when to stop the iterations.
`verbose`	If TRUE, the function prints the objective value on each run, which can be used to check if the objective is decreasing over iterations (as it should be) or not.

data("sample_by_feature_data")
out = Robocov_lowrank(sample_by_feature_data, stepsize = 1, max_iter = 100)
corrplot::corrplot(as.matrix(out$estR), 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")