wincrqa: Windowed Recurrence Measures

In crqa: Unidimensional and Multidimensional Methods for Recurrence Quantification Analysis

Windowed Recurrence Measures

Description

A recurrence plot is computed in overlapping windows of a certain size for a number of delays smaller than the size of the window; and measures of it extracted.

Usage

wincrqa(ts1, ts2, windowstep, windowsize, delay, embed, 
radius, rescale, normalize, mindiagline, minvertline, tw, whiteline, 
recpt, side, method, metric, datatype, trend)

Arguments

ts1	First time-series
ts2	Second time-series
windowstep	Interval by which the window is moved.
windowsize	The size of the window
delay	The delay unit by which the series are lagged.
embed	The number of embedding dimension for phase-reconstruction, i.e., the lag intervals.
radius	A threshold, cut-off, constant used to decide whether two points are recurrent or not.
rescale	Rescale the distance matrix; if rescale = 0 (do nothing); if rescale = 1 (mean distance of entire matrix); if rescale = 2 (maximum distance of entire matrix). if rescale = 3 (minimum distance of entire matrix). if rescale = 4 (euclidean distance of entire matrix).
normalize	Normalize the time-series; if normalize = 0 (do nothing); if normalize = 1 (Unit interval); if normalize = 2 (z-score).
mindiagline	A minimum diagonal length of recurrent points. Usually set to 2, as it takes a minimum of two points to define any line.
minvertline	A minimum vertical length of recurrent points.
tw	The Theiler window parameter
whiteline	A logical flag to calculate (TRUE) or not (FALSE) empty vertical lines.
recpt	A logical flag indicating whether measures of cross-recurrence are calculated directly from a recurrent plot (TRUE) or not (FALSE).
side	A string indicating whether recurrence measures should be calculated in the 'upper' triangle of the RP 'lower' triangle of the matrix, or 'both'. LOC is automatically excluded for 'upper' and 'lower'.
method	A string to indicate the type of recurrence analysis to peform. There are three options: rqa (autorecurrence); crqa(cross-recurrence); mdcrqa(multidimensional recurrence). Default value is crqa
metric	A string to indicate the type of distance metric used, default is euclidean but see help rdist() to list all other possible metrics.
datatype	a string (continuous or categorical) to indicate whether the nature of the data type
trend	a boolean (TRUE or FALSE) to indicate whether the TREND should be computed of the system

Value

It returns a matrix where the rows are the different windows explored, and the columns are the recurrence measures observed in that particular window. Refer to crqa for the values returned.

Note

If no-recurrence is found in a window, that window will not be saved, and a message about it will be warned. TREND is implemented following a solution proposed by Norbert Marwan, and translated here in R, for those who have asked him. He, however, warns that this measure might strongly depend on the chosen settings to calculate crq. Relying blindly on such measure can, therefore, produce misleading results. Also, we enabled the possibility to input directly a RP with recpt = T, and so extract the measures. This implies that it will not be the same as by conducting phase-space reconstruction on subsets of the time series. So, please, make sure why you are doing it and why.

Author(s)

Moreno I. Coco ([e-mail]) Alexandra Paxton [e-mail]

See Also

crqa

Examples

data(crqa) 

listener = eyemovement$listener[1:200]
narrator = eyemovement$narrator[1:200]

# NB, the parameters for windowsize and windowstep are large to allow
# faster running time, please set them carefully in your analysis. 

delay = 1; embed = 1; rescale = 0; radius = 0.001;
normalize = 0; mindiagline = 2; minvertline = 2;
tw = 0; whiteline = FALSE; recpt = FALSE; side = "both"
method = 'crqa'; metric = 'euclidean';  
datatype = "continuous"; 
windowsize =  100; windowstep = 50
trend = FALSE

ans = wincrqa(listener, narrator, windowstep, windowsize, delay, embed,
                    radius, rescale, normalize, mindiagline, minvertline,
                    tw, whiteline, recpt, side, method, metric, 
                    datatype, trend)

## other recurrence measures are available in ans
profile = as.numeric(ans$RR) 

plot(profile, type = 'l')

crqa documentation built on Nov. 27, 2023, 5:10 p.m.