Robust Range using Boxplot 'Quartiles'

Compute a robust range, i.e. the usual range() as long as there are no outliers, using the “whisker boundaries” of boxplot, i.e., boxplot.stats.

Usage

rrange(x, range=1, coef = 1.5, na.rm = TRUE)

Arguments

x

numeric vector the robust range of which shall be computed.

range

number for S compatibility; 1.5 * range is equivalent to coef.

coef

numeric multiplication factor definying the outlier boundary, see ‘Details’ below.

na.rm

logical indicating how NA values should be handled; they are simply dropped when na.rm = TRUE as by default.

Details

The robust range is really just what boxplot.stats(x, coef=coef) returns as the whisker boundaries. This is the most extreme values x[j] still inside median plus/minus coef * IQR.

Value

numeric vector c(m,M) with \(m \le M\) which is (not strictly) inside range(x) = c(min(x),max(x)).

Author

Martin Maechler, 1990.

See also

range, fivenum, boxplot and boxplot.stats.

A more sophisticated robust range for (strongly) asymmetric data can be derived from the skewness adjusted boxplot statistics adjboxStats which is a generalization of boxplot.stats.

Examples

stopifnot(rrange(c(1:10,1000)) == c(1,10))