capa.cc: Collective and point anomalies in cross-correlated...
In Tveten/capacc: Collective and point anomalies in cross-correlated data

Description Usage Arguments Value Examples

A method for detecting anomalous segments and points based on CAPA-CC by Tveten, Eckley, Fearnhead (2020).

capa.cc(
  x,
  Q,
  b = 1,
  b_point = 1,
  min_seg_len = 2,
  max_seg_len = 10^8,
  transform = centralise
)

`x`	An n x p data matrix where each row is an observation vector.
`Q`	An estimate of the precision matrix. See `robust_sparse_precision`. Must be a sparse matrix from the Matrix package.
`b`	The scaling factor for the collective anomaly penalty. Defaults to 1.
`b_point`	The scaling factor for the point anomaly penalty. Defaults to 1.
`min_seg_len`	The minimum segment length. Defaults to 2.
`max_seg_len`	The maximum segment length. Defaults to 10^8.
`transform`	A function used to centre the data prior to analysis by `capa.cc`. This can, for example, be used to compensate for the effects of autocorrelation in the data. The default value is `transform=centralise`, which centrailises the data by the median. Other choices are available in the anomaly package, for example `anomaly::robustscale` and `anomaly::ac_corrected`), and a user defined function can also be specified.

An S3 class of type capacc with the following components:

x: The input data matrix.
anoms: A data frame with four columns: start (start-point of the anomaly), end (end-point of an anomaly), variate (which variable is affected) and size (the estimated size of the mean component for the given variate).

library(capacc)
x <- simulate_cor()$x
Q <- robust_sparse_precision(x, adjacency_mat(banded_neighbours(2, ncol(x)), sparse = FALSE))
res <- capa.cc(x, Q, b = 1, min_seg_len = 5)
plot(res)
collective_anomalies(res)