boot_spec: Function that computes boostrapped confidence intervals for...

In timesboot: Bootstrap computations for time series objects

Description

The function resamples a time series object using phase scramble bootstrap and returns the average,lower and upper confidence bounds.

Usage

boot_spec(series, replic = 5000, spansa = c(11, 21), 
plot = TRUE, de_trend = FALSE, alpha = 0.05)

Arguments

series	A time series object
replic	The amount of replications used in the bootstrap step
spansa	The spans for smoothing the spectrum
plot	TRUE,FALSE. Whether plotting is desired. Default is TRUE.
de_trend	TRUE,FALSE. Should de-trending be applied to the series.Default is FALSE
alpha	The alpha used in the construction of the CI. Default is 0.05

Value

average	The average value for each frequency
upper	The upper value for each frequency
lower	The lower value for each frequency

Author(s)

Francisco Juretig

Examples

boot_spec(AirPassengers,replic=1000,alpha=0.05)

function (series, replic = 5000, spansa = c(11, 21), plot = TRUE, 
    de_trend = FALSE, alpha = 0.05) 
{
    if (is.ts(series) == TRUE) {
    	library(boot)
        kas = tsboot(series, redraw, p = spansa, detrend = de_trend, 
            replic, sim = "scramble")
        span1 = spansa[1]
        span2 = spansa[2]
        quantiles = matrix(0, length(kas$t[1, ]), 3)
        Xvalues = spec.pgram(series, spans = c(span1, span2), 
            plot = FALSE)
        for (i in 1:length(kas$t[1, ])) {
            cp = kas$t[, i]
            quantiles[i, 1] = quantile(cp, alpha)
            quantiles[i, 2] = quantile(cp, 1 - alpha/2)
            quantiles[i, 3] = mean(cp)
        }
        if (plot == TRUE) {
            maximus = max(quantiles[, 3])
            plot(Xvalues$freq, quantiles[, 1], type = "l", col = "blue", 
                ylim = c(0, maximus * 2), main = "Bootstraped Periodogram", 
                ylab = "periodogram", lwd = 1, xlab = "frequency")
            polygon(c(Xvalues$freq, rev(Xvalues$freq)), c(quantiles[, 
                1], rev(quantiles[, 2])), col = "skyblue")
            lines(Xvalues$freq, quantiles[, 3], type = "o", col = "black", 
                pch = 20)
        }
        lix = list(freq = Xvalues$freq, upper = quantiles[, 2], 
            lower = quantiles[, 1], mean = quantiles[, 3])
        return(lix)
    }
    else {
        return("Object is not a time-series")
    }
  }

timesboot documentation built on May 2, 2019, 2:37 p.m.