mark_outliers: mark_outliers (package "deptstat")

In mithridata-com/deptstat: Data processing and filtering

Description

This function is created to mark outliers in any selected column of a dataframe. The idea is based on Tukey's method (Tukey, 1977) for outlier trimming: The lower quartile (q1) is the 25th percentile, and the upper quartile (q3) is the 75th percentile of the data. The inter-quartile range (IQR) is defined as the interval between q1 and q3. Tukey (1997) defined q1-(1.5*iqr) and q3+(1.5*iqr) as 'inner fences'. So called 'outer fences' are not supported yet. Function return a vector (or a column in a data frame) that contains resulting value of outliers: 0 - ordinary observation, 1 - outlier. Hence, the function supports alternative options like one-sided trimming ("right", "left") or different quantiles of distribution (default: 0.25 and 0.75)

Usage

mark_outliers(
  data,
  var = "",
  tukey.coef = 1.5,
  trim.type = "two-sided",
  quantiles = c(0.25, 0.75),
  verbose = FALSE,
  return_df = FALSE
)

Arguments

data	Ordinary R data frame with at least one numeric column
var	A particular numeric column of a data frame, that will be examined for outliers. Default:"" (nothing)
tukey.coef	A coefficient is defaulted to be 1.5 but can me modified (for example, Tukey used coefficient 1.5 for inner fence and 3 for outer fences). Default: 1.5
trim.type	A type of trimming: "two-sided" - when you search for outliers on both sides of a distribution, "right" - when you search for outlies on the right side of a distibution, "left" - when search for outliers on the left side of a distribution. Default:"two-sided"
quantiles	A vector of two values with percentiles that will be used for calculation of IQR and to form fences (quantiles +/- tukey.coef * IQR). Default: c(0.25, 0.75)
verbose	A verbose parameter setted to TRUE will print additional summary output. Default: FALSE
return_df	A parameter setted to TRUE will return a data frame (instead of vector) with input column, marked outliers and extra data on IQR, tukey.coef and quantiles. Default: FALSE

Value

A vector or data frame with outliers marked as 1 and normal observations marked as 0

Author(s)

Dmitrii Diachkov (2021)

Examples

# Let's find outliers in mtcars dataset
data("mtcars")
# Let's examine dataset first to find out variable names
summary(mtcars)
# Now we try out luck in outliers with mpg column (Miles/(US) gallon)
# and save a result into a separate dataset
mtcars_with_outliers <- mark_outliers(mtcars, "mpg", return_df = T)
# If you inspect mtcars_with_outliers, you will find out that Toyota Corolla
# is marked as and outlier, because it's mpg is 33.9
# while right fence is 22.8+1.5*7.375=33.8625
# Let's plot data with outliers
plot(mtcars_with_outliers[["mpg"]], main = "Marked outliers",
     col = factor(mtcars_with_outliers$mpg.is_outlier), pch = 19)
legend("bottomright", legend = c("Normal data","Outlier"), col = 1:2, pch = 19, bty = "n")

mithridata-com/deptstat documentation built on Dec. 21, 2021, 7:01 p.m.