Rle-runstat: Fixed-width running window summaries

In Bioconductor/S4Vectors: Foundation of vector-like and list-like containers in Bioconductor

Fixed-width running window summaries

Description

The runsum, runmean, runmed, runwtsum, runq functions calculate the sum, mean, median, weighted sum, and order statistic for fixed width running windows.

Usage

runsum(x, k, endrule = c("drop", "constant"), ...)

runmean(x, k, endrule = c("drop", "constant"), ...)

## S4 method for signature 'Rle'
smoothEnds(y, k = 3)

## S4 method for signature 'Rle'
runmed(x, k, endrule = c("median", "keep", "drop", "constant"), 
       algorithm = NULL, print.level = 0)

runwtsum(x, k, wt, endrule = c("drop", "constant"), ...)

runq(x, k, i, endrule = c("drop", "constant"), ...)

Arguments

x, y	The data object.
k	An integer indicating the fixed width of the running window. Must be odd when endrule != "drop".
endrule	A character string indicating how the values at the beginning and the end (of the data) should be treated. "median" see runmed; "keep" see runmed; "drop" do not extend the running statistics to be the same length as the underlying vectors; "constant" copies running statistic to the first values and analogously for the last ones making the smoothed ends constant.
wt	A numeric vector of length k that provides the weights to use.
i	An integer in [0, k] indicating which order statistic to calculate.
algorithm, print.level	See ?stats::runmed for a description of these arguments.
...	Additional arguments passed to methods. Specifically, na.rm. When na.rm = TRUE, the NA and NaN values are removed. When na.rm = FALSE, NA is returned if either NA or NaN are in the specified window.

Details

The runsum, runmean, runmed, runwtsum, and runq functions provide efficient methods for calculating the specified numeric summary by performing the looping in compiled code.

Value

An object of the same class as x.

Author(s)

P. Aboyoun and V. Obenchain

See Also

runmed, Rle-class, RleList-class

Examples

  x <- Rle(1:10, 1:10)
  runsum(x, k = 3)
  runsum(x, k = 3, endrule = "constant")
  runmean(x, k = 3)
  runwtsum(x, k = 3, wt = c(0.25, 0.5, 0.25))
  runq(x, k = 5, i = 3, endrule = "constant")

  ## Missing and non-finite values
  x <- Rle(c(1, 2, NA, 0, 3, Inf, 4, NaN))
  runsum(x, k = 2)
  runsum(x, k = 2, na.rm = TRUE)
  runmean(x, k = 2, na.rm = TRUE)
  runwtsum(x, k = 2, wt =  c(0.25, 0.5), na.rm = TRUE)
  runq(x, k = 2, i = 2, na.rm = TRUE) ## max value in window

  ## The .naive_runsum() function demonstrates the semantics of 
  ## runsum(). This test ensures the behavior is consistent with
  ## base::sum().

  .naive_runsum <- function(x, k, na.rm=FALSE)
      sapply(0:(length(x)-k),
          function(offset) sum(x[1:k + offset], na.rm=na.rm))

  x0 <- c(1, Inf, 3, 4, 5, NA)
  x <- Rle(x0)
  target1 <- .naive_runsum(x0, 3, na.rm = TRUE)
  target2 <- .naive_runsum(x, 3, na.rm = TRUE)
  stopifnot(target1 == target2)
  current <- as.vector(runsum(x, 3, na.rm = TRUE))
  stopifnot(target1 == current)

  ## runmean() and runwtsum() :
  x <- Rle(c(2, 1, NA, 0, 1, -Inf))
  runmean(x, k = 3)
  runmean(x, k = 3, na.rm = TRUE)
  runwtsum(x, k = 3, wt = c(0.25, 0.50, 0.25))
  runwtsum(x, k = 3, wt = c(0.25, 0.50, 0.25), na.rm = TRUE)

  ## runq() :
  runq(x, k = 3, i = 1, na.rm = TRUE) ## smallest value in window
  runq(x, k = 3, i = 3, na.rm = TRUE) ## largest value in window

  ## When na.rm = TRUE, it is possible the number of non-NA 
  ## values in the window will be less than the 'i' specified.
  ## Here we request the 4th smallest value in the window,
  ## which tranlates to the value at the 4/5 (0.8) percentile.
  x <- Rle(c(1, 2, 3, 4, 5))
  runq(x, k=length(x), i=4, na.rm=TRUE)

  ## The same request on a Rle with two missing values
  ## finds the value at the 0.8 percentile of the vector 
  ## at the new length of 3 after the NA's have been removed.
  ## This translates to round((0.8) * 3).
  x <- Rle(c(1, 2, 3, NA, NA))
  runq(x, k=length(x), i=4, na.rm=TRUE)

Bioconductor/S4Vectors documentation built on Nov. 17, 2024, 6:55 p.m.