bootstrap_C: Bootstrap a cutoff value to identify anomalies

In cshannum/unequalgroupoutlier: Anomaly Detection in Unequal Length Functional Data (Title Case)

Description

Bootstrap a cutoff value to identify anomalies

Usage

bootstrap_C(coeff, d.method, c.method, alpha, B)

Arguments

coeff	A dataframe of coefficients of interest. The first column is ID identifier. The rest of the columns are for the parameter to be estimate. Each row is the estimated parameters fore each curve.
d.method	A character string determining the depth function to use: "LP", "Projection", "Mahalanobis", or "Euclidean". It is suggested to not use "Tukey" due to singularity in coefficient matrix. For details see depth
c.method	A character string determining the method to estimate the cutoff value. This can be "depth" or "alpha".
alpha	A value determining the percentage of rows to remove from coeff. alpha should be between (0, 1) with a suggested value of 0.05. Do not need to identify if c.method = "depth".
B	A value determining how many bootstrap datasets should be made to estimate the cutoff value with a suggested rate of 1000.

Details

The function starts by computing the depths for each parameter set by d.method.

The "alpha" c.method removes the alpha percent least deep coefficients. The rest of the coefficients are bootstrapped and new depths are computed for each new bootstrapped set. The 1 1

The "depth" c.method bootstraps the coefficients with probability related to the original depth values. New depths are computed for each new bootstrapped set. The 1 1

Value

$d the depths computed by d.method over all coefficients. $Cb the cutoff value; depths below cutoff may be anomalous.

See Also

depth, bootstrap_C.alpha, and bootstrap_C.depth

cshannum/unequalgroupoutlier documentation built on May 13, 2019, 11:10 a.m.