bumphunter (version 1.12.0)

getSegments: Segment a vector into positive, zero, and negative regions

Description

Given two cutoffs, L and U, this function divides a numerical vector into contiguous parts that are above U, between L and U, and below L.

Usage

getSegments(x, f = NULL, cutoff = quantile(abs(x), 0.99), assumeSorted = FALSE, verbose = FALSE)

Arguments

x

A numeric vector.

f

A factor used to pre-divide x into pieces. Each piece is then segmented based on the cutoff. Setting this to NULL says that there is no pre-division. Often, clusterMaker is used to define this factor.

cutoff

a numeric vector of length either 1 or 2. If length is 1, U (see details) will be cutoff and L will be -cutoff. Otherwise it specifies L and U. The function will furthermore always use the minimum of cutoff for L and the maximum for U.

assumeSorted

This is a statement that the function may assume that the vector f is sorted. Allowing the function to make this assumption may increase the speed of the function slightly.

verbose

Should the function be verbose?

Value

A list with three components, each a list of indices. Each component of these lists represents a segment and this segment is represented by a vector of indices into the original vectors x and f.

upIndex:

a list with each entry an index of contiguous values in the same segment. These segments have values of x above U.

dnIndex:

a list with each entry an index of contiguous values in the same segment. These segments have values of x below L.

zeroIndex:

a list with each entry an index of contiguous values in the same segment. These segments have values of x between L and U.

Details

This function is used to find the indexes of the ‘bumps’ in functions such as bumphunter.

x is a numeric vector, which is converted into three levels depending on whether x>=U (‘up’), Lcutoff. We assume that adjacent entries in x are next to each other in some sense. Segments, consisting of consecutive indices into x (ie. values between 1 and length(x)), are formed such that all indices in the same segment belong to the same level of f and have the same discretized value of x.

In other words, we can use getSegments to find runs of x belonging to the same level of f and with all of the values of x either above U, between L and U, or below L.

See Also

clusterMaker

Examples

x <- 1:100
y <- sin(8*pi*x/100)
chr <- rep(1, length(x))
indexes <- getSegments(y, chr, cutoff=0.8)
plot(x, y, type="n")
for(i in 1:3){
   ind <- indexes[[i]]
   for(j in seq(along=ind)) {
      k <- ind[[j]]
      text(x[k], y[k], j, col=i)
   }
}
abline(h=c(-0.8,0.8))