Outlier Alert for Numeric Vectors

This helper function identifies numeric outliers based on the bounds computed by LocScaleB. It optionally applies a log transform within LocScaleB (through the ... argument) and flags values outside the computed lower or upper bounds. Values below the lower bound receive the alert code alert_if_smaller, values above the upper bound receive alert_if_larger, and in-range values receive no_alert_value.

Usage

alert_outlier(
  x,
  no_alert_value = NA_real_,
  alert_if_larger = no_alert_value,
  alert_if_smaller = no_alert_value,
  ...
)

Arguments

x

A numeric vector in which to detect outliers.

no_alert_value

A numeric code (default NA_real_) to assign to non-outlier values.

alert_if_larger

A numeric code to assign if x > upper bound.

alert_if_smaller

A numeric code to assign if x < lower bound.

...

Additional arguments passed to LocScaleB, such as logt or k.

Value

A numeric vector of the same length as x, containing the alert codes for each element (alert_if_smaller, alert_if_larger, or no_alert_value).

Details

• The function checks if all x values are NA or zero, or if the MAD (median absolute deviation) is zero. In these cases, it returns NA for all elements, since meaningful bounds cannot be computed.

• Otherwise, it calls univOutl::LocScaleB(x, ...) to compute bounds.

• If a log transform was used (logt = TRUE), LocScaleB returns the log-scale bounds; typically, you would back-transform them to compare with the raw values.

See also

LocScaleB for the underlying outlier detection algorithm.

Examples

if (FALSE) { # \dontrun{
x <- c(1, 2, 3, 100)
alert_outlier(x, no_alert_value = NA, alert_if_larger = 9, logt = TRUE, k = 3)
} # }