roll_hampel: Rolling Hampel Filter for Outlier Detection

In seismicRoll: Fast Rolling Functions for Seismology using 'Rcpp'

Rolling Hampel Filter for Outlier Detection

Description

Fast, center-aligned hampel filter using C++/Rcpp. The Hampel filter is a robust outlier detector using Median Absolute Deviation (MAD). Additional performance gains can be achieved by skipping increment values between calculations.

Usage

roll_hampel(x, n, increment = 1)

Arguments

x	an R numeric vector
n	integer window size
increment	integer shift to use when sliding the window to the next location

Details

Unlike the version in the pracma package, this version does not return the corrected timeseries. Instead, it returns a vector of values that can be tested against different threshold values. Higher values in the return are associated with a higher likelihood that the associated point is an outlier when compared with its neighbors. Outliers can be picked out by comparing the return values against some threshold as seen in the example.

Also unlike the pracma version, n is interpreted as the full window length and will be increased by one if necessary to have a window representing an odd number of indices.

Value

A vector of values of the same length as x.

Note

A pure R version of the filter is found in the pracma package.

See Also

roll_median

Examples

a <- sin(0.1*seq(100))
a[20] <- 50

b <- roll_hampel(a,10)
threshold <- 6
which(b > threshold)

## Not run: 
  require(microbenchmark)
  require(pracma)
  
  microbenchmark(hampel(a,10), roll_hampel(a,10), times=10)
  
  #  Unit: microseconds
  #                 expr      min       lq    median       uq       max neval
  #        hampel(a, 10) 7610.688 7784.374 8037.4035 9453.928 16176.535    10
  #   roll_hampel(a, 10)   36.530   37.443   58.7165   65.418    90.403    10

## End(Not run)

seismicRoll documentation built on Aug. 22, 2023, 9:13 a.m.