exclude_outliers: Exclude outliers from a data set

Description

Excludes outliers from a data set using the "1.5 interquartile range" rule.

Usage

exclude_outliers(x, col_for_analysis, INDICES, method = 'exclude')
  # S3 method for data.frame
exclude_outliers(x, col_for_analysis, INDICES, method = 'exclude')
  # S3 method for exdf
exclude_outliers(x, col_for_analysis, INDICES, method = 'exclude')

Value

This function returns an object formed from x, where the results depend on on the value of method.

When method is 'remove', the returned object is a modified copy of x where all rows in which the value of col_for_analysis is an outlier have been removed.

When method is 'exclude', the returned object is a modified copy of x where all outlier values of col_for_analysis have been replaced with NA.

Arguments

x: A data table
col_for_analysis: The name of a column of x that should be used to determine outliers.
INDICES: A factor or list of factors that each nrow(x) elements.
method: Specify whether to remove rows from x ('remove') or to replace outlier values of col_for_analysis with NA ('exclude').

Details

exclude_outliers is generic, with methods defined for data frames and exdf objects. This function uses a simple rule to detect outliers, where any point that deviates from the mean by more than 1.5 * IQR, where IQR is the interquartile range, is said to be an outlier. This method is also sometimes referred to as "Tukey's Fences," as seen in the Wikipedia page about outliers.

For data sets with extreme outliers, it may be necessary to exclude outliers more than once to actually remove them all.

Examples

Run this code

# Read a Licor file included with the PhotoGEA package; this file includes
# several light response curves that can be identified by the 'species' and
# 'plot' columns.
licor_file <- read_gasex_file(
  PhotoGEA_example_file_path('ball_berry_1.xlsx')
)

# Remove points from each response curve in the data where the leaf temperature
# is determined to be an outlier
licor_file_clean <- exclude_outliers(
  licor_file,
  'TleafCnd',
  list(licor_file[, 'species'], licor_file[, 'plot']),
  method = 'remove'
)

# Check to see how many points remain after removing outliers
str(list('original' = nrow(licor_file), 'clean' = nrow(licor_file_clean)))