Robust smoothing by running medians

A sequence of running-median smoothing passes, resistant to outliers the way a mean is not.

Usage

tda_smd(x, sm = "3R", options = list(), dir = tempfile("tda"))

Arguments

x

the variable to smooth.

sm

the smoothing operations as a string of characters: "3" is a running median of three, "3R" repeats it to convergence, "2" and "4" are medians of two and four, and they are concatenated to apply in sequence, so "33" is two passes of a 3-median. Not a vector of numbers – a comma is a syntax error.

options

a named list of further TDA options, passed through.

dir

working directory.

Value

An object carrying a table: one row per case, with the raw value beside its smoothed one.

See also

Other smoothing: tda_integrate(), tda_interp(), tda_isotonic(), tda_mat(), tda_sma(), tda_spl()

Examples

# a running median is robust to the single outlier at position 9
x <- c(5, 2, 8, 3, 9, 1, 7, 4, 100, 6)
tda_smd(x, sm = "3R")$table
#>    index case raw smoothed
#> 1      0    1   5        5
#> 2      0    2   2        5
#> 3      0    3   8        5
#> 4      0    4   3        5
#> 5      0    5   9        5
#> 6      0    6   1        6
#> 7      0    7   7        6
#> 8      0    8   4        6
#> 9      0    9 100        6
#> 10     0   10   6        6