dist_simu_test: Cut-Off Values Using Simulations for the Detection of Extreme...

In ICSOutlier: Outlier Detection Using Invariant Coordinate Selection

Cut-Off Values Using Simulations for the Detection of Extreme ICS Distances

Description

Computes the cut-off values for the identification of the outliers based on the squared ICS distances. It uses simulations under a multivariate standard normal model for a specific data setup and scatters combination.

Usage

dist_simu_test(
  object,
  S1 = NULL,
  S2 = NULL,
  S1_args = list(),
  S2_args = list(),
  index,
  m = 10000,
  level = 0.025,
  n_cores = NULL,
  iseed = NULL,
  pkg = "ICSOutlier",
  q_type = 7,
  ...
)

Arguments

object	object of class "ICS" where both S1 and S2 are specified as functions. The sample size and the dimension of interest are also obtained from the object. The invariant coordinate are required to be centered.
S1	an object of class "ICS_scatter" or a function that contains the location vector and scatter matrix as location and scatter components.
S2	an object of class "ICS_scatter" or a function that contains the location vector and scatter matrix as location and scatter components.
S1_args	a list containing additional arguments for S1.
S2_args	a list containing additional arguments for S2.
index	integer vector specifying which components are used to compute the ics_distances().
m	number of simulations. Note that extreme quantiles are of interest and hence m should be large.
level	the (1-level(s))th quantile(s) used to choose the cut-off value(s). Usually just one number between 0 and 1. However a vector is also possible.
n_cores	number of cores to be used. If NULL or 1, no parallel computing is used. Otherwise makeCluster with type = "PSOCK" is used.
iseed	If parallel computation is used the seed passed on to clusterSetRNGStream. Default is NULL which means no fixed seed is used.
pkg	When using parallel computing, a character vector listing all the packages which need to be loaded on the different cores via require. Must be at least "ICSOutlier" and must contain the packages needed to compute the scatter matrices.
q_type	specifies the quantile algorithm used in quantile.
...	further arguments passed on to the function quantile.

Details

The function extracts basically the dimension of the data from the "ICS" object and simulates m times, from a multivariate standard normal distribution, the squared ICS distances with the components specified in index. The resulting value is then the mean of the m correponding quantiles of these distances at level 1-level.

Note that depending on the data size and scatters used this can take a while and so it is more efficient to parallelize computations.

Note that the function is seldomly called directly by the user but internally by ICS_outlier().

Value

A vector with the values of the (1-level)th quantile.

Author(s)

Aurore Archimbaud and Klaus Nordhausen

References

Archimbaud, A., Nordhausen, K. and Ruiz-Gazen, A. (2018), ICS for multivariate outlier detection with application to quality control. Computational Statistics & Data Analysis, 128:184-199. ISSN 0167-9473. \Sexpr[results=rd]{tools:::Rd_expr_doi("10.1016/j.csda.2018.06.011")}.

See Also

ICS(), ics_distances()

Examples

# For a real analysis use larger values for m and more cores if available

Z <- rmvnorm(1000, rep(0, 6))
Z[1:20, 1] <- Z[1:20, 1] + 10
A <- matrix(rnorm(36), ncol = 6)
X <- tcrossprod(Z, A)

pairs(X)
icsX <- ICS(X, center = TRUE)

icsX.dist.1 <- ics_distances(icsX, index = 1)
CutOff <- dist_simu_test(icsX, S1 = ICS_cov, S2= ICS_cov4,
                        index = 1, m = 500, ncores = 1)

# check if outliers are above the cut-off value
plot(icsX.dist.1, col = rep(2:1, c(20, 980)))
abline(h = CutOff)


library(REPPlab)
data(ReliabilityData)
# The observations 414 and 512 are suspected to be outliers
icsReliability <- ICS(ReliabilityData, center = TRUE)
# Choice of the number of components with the screeplot: 2
screeplot(icsReliability)
# Computation of the distances with the first 2 components
ics.dist.scree <- ics_distances(icsReliability, index = 1:2)
# Computation of the cut-off of the distances 
CutOff <- dist_simu_test(icsReliability, S1 = ICS_cov, S2= ICS_cov4,
                         index = 1:2, m = 50, level = 0.02, ncores = 1)
# Identification of the outliers based on the cut-off value
plot(ics.dist.scree)
abline(h = CutOff)
outliers <- which(ics.dist.scree >= CutOff)
text(outliers, ics.dist.scree[outliers], outliers, pos = 2, cex = 0.9)

## Not run: 
    # For using three cores
    # For demo purpose only small m value, should select the first #' component
    dist_simu_test(icsReliability, S1 = ICS_cov, S2= ICS_cov4,
                   index = 1:2, m = 500, level = 0.02, n_cores = 3, iseed #' = 123)
    
    # For using several cores and for using a scatter function from a different package
    # Using the parallel package to detect automatically the number of cores
    library(parallel)
    # ICS with Cauchy estimates
    library(ICSClust)
    icsReliabilityMLC <- ICS(ReliabilityData, S1 = ICS_mlc, 
                            S1_args = list(location = TRUE),
                                 S2 = ICS_cov, center = TRUE)
    # Computation of the cut-off of the distances. For demo purpose only small m value.
    dist_simu_test(icsReliabilityMLC,  S1 = ICS_mlc,  S1_args = list(location = TRUE), 
    S2 = ICS_cov, index = 1:2, m = 500, level = 0.02, 
    n_cores = detectCores()-1,  pkg = c("ICSOutlier","ICSClust"), iseed = 123)

## End(Not run)

ICSOutlier documentation built on May 29, 2024, 2:08 a.m.