Ditrack: Class methods for Ditrack Objects

In Dimodal: Spacing Tests for Multi-Modality

Class methods for Ditrack Objects

Description

Find peaks and flats that appear in the spacing as the size of the low-pass filter or interval changes. Support methods for the class include printing and plotting the significant features.

Usage

Ditrack(x, smooth=c('lp', 'diw'), opt=Diopt())
## S3 method for class 'Ditrack'
print(x, ...)
## S3 method for class 'Ditrack'
summary(object, ...)
## S3 method for class 'Ditrack'
plot(x, feature=c('peaks', 'flats'), opt=Diopt(), ...)

Arguments

x	for Ditrack, the data to study; for the class methods, the value returned from Ditrack
object	an object of class "Ditrack"
smooth	which filtering to apply to the data, "lp" to look for features in the low-pass spacing or "diw" in the interval spacing
opt	local version of options passed to Dimodal analysis
feature	which feature(s) to plot, may include both
...	extra arguments, ignored for all methods

Details

The mode tree and SiZer visualization tools track a feature's position and some metric as the bandwidth of a kernel density estimate changes. They help find the best bandwidth for the trade-off between feature detectability and smoothing. The Ditrack class does the same for the spacing analysis. It varies the size of the low-pass filter kernel or interval spacing and tracks the position and probability of peaks and flats per the tests in the Dimodal analysis.

The Ditrack function varies the window size and gathers the peak and flat results into matrices.

The tracker will overwrite the options ⁠"analysis"⁠ and either ⁠"lp.window"⁠ or ⁠"diw.window"⁠ from Diopt. These should not appear in the ⁠"lp.param"⁠ or ⁠"diw.param"⁠ overrides, which have precedence. It steps these sizes from 0.01 to the option ⁠"track.maxwindow"⁠.

plot.Ditrack graphs one or both features, with the window size on the vertical axis and the position along the horizontal. Peaks and minima are drawn as symbols, with dots for the maxima and dashes for the minima. The dots are filled in if any test passes its acceptance level. Flats are drawn with a line spanning the feature. Both are color-coded according to the overall probability, the best of all tests that were run. The coloring is fixed and does not use the ⁠"palette"⁠ option. There will always be a minimum between two peaks. Extrema at the edges of the data are not drawn.

print.Ditrack gives a table with the number of significant features at the 0.01 and 0.05 level plus the number of features significant according to the current acceptance levels, as the window size changes.

The summary method prints the ranges of window sizes for peaks and flats that give features that pass the acceptance levels of any test that has been run.

Value

The tracker returns a list assigned to class ⁠"Ditrack"⁠ with elements

peaks	a subset of the ⁠"Dipeak"⁠ result with columns ⁠"pos"⁠, ⁠"ismax"⁠, ⁠"ppeak"⁠, and ⁠"naccept"⁠, plus an additional column ⁠"winpct"⁠ giving the filter size as an integer percentage, keeping only rows with valid peaks
flats	a subset of the ⁠"Diflat"⁠ result taking columns ⁠"stID"⁠, ⁠"endID"⁠, ⁠"pflat"⁠, and ⁠"naccept"⁠, plus an additional column ⁠"winpct"⁠ giving the filter size
nx	the length of the data x argument
smooth	the argument of the same name

The peaks and flats matrices will have one row per feature, be it a minimum or maximum from Dipeak or a flat. There may be many rows with the same winpct, or none for a given value.

See Also

Dimodal, Dipeak, Diflat

Examples

  ## Not run because of the run time.
   trk <- Ditrack(quakes[,3], 'lp') 
  ## Two plots side by side.
   dev.new(width=8,height=4) ; plot(trk) 
   trk 

Dimodal documentation built on May 2, 2026, 1:06 a.m.