outlier_tukey: Flag outliers based on tukey's rule

In southwick-associates/sastats: Calculate Statistics

Description

Returns a logical vector where TRUE indicates outliers.

Usage

outlier_tukey(
  x,
  k = 1.5,
  ignore_lwr = FALSE,
  apply_log = FALSE,
  ignore_zero = FALSE
)

outlier_tukey_top(x, k = 1.5, apply_log = FALSE, ignore_zero = FALSE)

Arguments

x	input values to check
k	the iqr multiplier that determines the fence level. Increasing will make outlier identification less strict (& vice-versa)
ignore_lwr	If TRUE, don't use the lower fence for identifying outliers
apply_log	If TRUE, log transform input values prior to applying tukey's rule. Useful since distributions often have a log-normal shape (e.g., spending)
ignore_zero	If TRUE, will exclude zero values from IQR & flagging. Note that zeroes will automatically be ignored if apply_log = TRUE

Functions

• outlier_tukey_top: get the largest non-outlier value for top-coding

See Also

Other functions for identifying outliers: outlier_mean_compare(), outlier_pct(), outlier_plot()

Examples

library(dplyr)
data(svy)

# take a look at the days variable
outlier_plot(svy$act, days, act)
outlier_plot(svy$act, days, act, apply_log = TRUE)

activity <- group_by(svy$act, act) %>% mutate(
    is_outlier = outlier_tukey(days, ignore_zero = TRUE, apply_log = TRUE),
    # in case we want to topcode the outliers:
    topcode_value = outlier_tukey_top(days, apply_log = TRUE),
    days_cleaned = ifelse(is_outlier, NA, days)
) %>% ungroup()

# summarize
outlier_plot(activity, days, act, apply_log = TRUE, show_outliers = TRUE)
outlier_pct(activity, act)
outlier_mean_compare(activity, days, days_cleaned, act)

southwick-associates/sastats documentation built on March 27, 2020, 9:39 p.m.