Free and open-source implementation in R of Outskewer, a method to detect outliers in data sets of numbers and in time series. It is distributed under the MIT License.

###Usage ####Load functions

setwd('path/to/outskewer')  # to edit
source("outskewer.r")
source("utils.r")

####Execute on a vector of numbers

tmp <- DetectOutliers(x)
results.df <- SetFinalOutlierStatus(tmp$df)

####Execute on a time series

# Load data, give proper names to the data frame columns.
# File structure: timestamp, value
# 1 10807
# 2 10873
# 3 10902
data.df <- read.table('ts.dat')
colnames(data.df) <- c("t","x")

# Run outskewer with a time window with size of w values.
results.df <- DetectOutliersDynamically(data.df, w=100)
results.df <- SetFinalOutlierStatus(results.df)

###Display results ####Number of outliers detected

summary(factor(results.df$status))

####Cumulative frequency with points colored by outlier status

g <- PlotCumulative(results.df)
g <- DecoratePlot(g)  # optional
g

####Skewness signature of a distribution of values

PlotSignature(x)

####Time series with points colored by outlier status

g <- PlotTimeSeries(results.df)
g <- DecoratePlot(g)  # optional
g

###Save in file ####Results

# Save in a tabular format
ExportData(data=results.df, filename='my_results')

####Plot

# Save in PDF, EPS and PNG 
ExportPlot(gplot=g, filename='my_results', width=10, height=5)