plot.locOuts {ssMRCD}

Diagnostic Plots for Local Outlier Detection

Description

This function plots different diagnostic plots for local outlier detection. It can be applied to an object of class "locOuts" which is the output of the function local_outliers_ssMRCD.

Usage

## S3 method for class 'locOuts'
plot(
  x,
  type = c("hist", "spatial", "lines", "3D"),
  colour = "all",
  focus = NULL,
  pos = NULL,
  alpha = 0.3,
  data = NULL,
  add_map = TRUE,
  ...
)

Arguments

x

a locOuts object obtained by the function local_outliers_ssMRCD.

type

vector containing the types of plots that should be plotted, possible values c("hist", "spatial", "lines", "3D").

colour

character specifying the color scheme (see details). Possible values "all", "onlyOuts", "outScore".

focus

an integer being the index of the observation whose neighborhood should be analysed more closely.

pos

integer specifying the position of the text "cut-off" in the histogram (see par).

alpha

scalar specifying the transparancy level of the points plotted for plot type "spatial", "3D" and "lines".

data

optional data frame or matrix used for plot of type "line". Will be used to plot lines based scaled data instead of the data used for local outlier detection.

add_map

TRUE if a map should be plotted along the line plot (type = "lines").

...

further parameters passed on to base-R plotting functions.

Details

Regarding the parameter type the value "hist" corresponds to a plot of the histogram of the next distances together with the used cutoff-value. When using "spatial" the coordinates of each observation are plotted and colorized according to the color setting. The "lines" plot is used with the index focus of one observation whose out/inlyingness to its neighborhood should by plotted. The whole data set is scaled to the range [0,1] and the scaled value of the selected observation and its neighbors are plotted. Outliers are plotted in orange. The "3D" setting leads to a 3D-plot using the colour setting as height. The view can be adapted using the parameters theta and phi.

For the colour setting possible values are "all" (all next distances are used and colored in an orange palette), "onlyOuts" (only outliers are plotted in orange, inliers are plotted in grey) and "outScore" (the next distance divided by the cutoff value is used to colourize the points; inliers are colorized in blue, outliers in orange).

Value

Returns plots regarding next distances and spatial context.

See Also

local_outliers_ssMRCD

Examples

# set seed
set.seed(1)

# make locOuts object
data = matrix(rnorm(2000), ncol = 4)
coords = matrix(rnorm(1000), ncol = 2)
N_assignments = sample(1:10, 500, replace = TRUE)
lambda = 0.3

# local outlier detection
outs = local_outliers_ssMRCD(data = data,
                             coords = coords,
                             N_assignments = N_assignments,
                             lambda = lambda,
                             k = 10)

# plot results
plot(outs, type = "hist")
plot(outs, type = "spatial", colour = "outScore")
plot(outs, type = "3D", colour = "outScore", theta = 0)
plot(outs, type ="lines", focus = outs$outliers[1])