Trimmed Mean (robust location)

Computes the trimmed mean of a one-dimensional signal – the arithmetic mean after discarding the most extreme trim fraction of the values from each end. A robust measure of central tendency that, unlike the ordinary mean, is not pulled by a heavy tail or a few outliers, while (unlike the median) it still uses all of the retained data. It matches base R mean(x, trim = ...) and scipy.stats.trim_mean bit-for-bit, both of which discard \(\lfloor n\,\mathrm{trim}\rfloor\) values from each end.

Usage

trimmedMean(x, trim = 0.1)

Arguments

x

A numeric vector (the time series).

trim

Fraction (0 to 0.5) of values trimmed from EACH end (default 0.1).

Value

A single numeric value, the trimmed mean.

Details

A robust location estimator that completes the robust toolkit alongside medianAbsDev and interquartileRange (robust dispersion). On a skewed distribution the trimmed mean lies between the ordinary mean and the median, and moves toward the median as trim increases – a measure of how much the tail inflates the mean.

See also

signalMoments for the ordinary mean and SD, medianAbsDev and interquartileRange for robust dispersion.

Examples

trimmedMean(c(1, 2, 3, 4, 100), trim = 0.2)  # 3 (robust); the mean is 22
#> [1] 3
set.seed(1); trimmedMean(rexp(1000), trim = 0.2)  # below the tail-inflated mean
#> [1] 0.8067975