sour: Cross-correlation of time series (which may be unevenly sampled)

Description Usage Arguments Details Value Notes See Also Examples

fr_rss returns a randomise version of an input data array.

1	fr_rss(dat)

dat

(data frame) containing columns t, y and (optionally) dy.

Performs "flux randomisation" and "random sample selection" of an input time series, following Peterson et al. (2004, ApJ, v613, pp682-699). This is essentially a bootstrap for a data vector.

A data frame containing columns

`t`	time bins for randomised data
`y`	values for randomised data
`dy`	errors for randomised data

Given an input data series (t, y, dy) of length N we sample N points with replacement. Duplicated points are ignored, so the ouptut is usually shorter than the input. So far this is a basic bootstrap procedure.

If error bars are provided: when a point is selected m times, we decrease the error, scaling by 1/sqrt(m). See Appendix A of Peterson et al. After resampling, we then add a random Gaussian deviate to each remaining data point, with std.dev equal to its (new) error bar. If errors bars are not provided, this is a simple bootstrap (no randomisation of y).

cross_correlate, ccf_errors

 ## Example using the NGC 5548 data
 plot(cont$t, cont$y, type="l", bty = "n", xlim = c(50500, 52000))
 rcont <- fr_rss(cont)
 lines(rcont$t, rcont$y, col = "red")

 ## Examples from Venables & Ripley
 require(graphics)
 plot(fdeaths, bty = "n")
 tsf <- data.frame(t = time(fdeaths), y = fdeaths)
 rtsf <- fr_rss(tsf)
 lines(rtsf$t, rtsf$y, col="red", type="o")