boot_farr_fit.np: Non-parametric estimation of the far from bootstrap samples
In thaos/farr: Computation of the Fraction of Attributable Risk for Records

Description Usage Arguments Details Value Examples

boot_farr_fit.np returns an object of class ("boot_farrfit.np", "boot_farrfit", "farrfit") which contains the estimates of the far for each bootstrap sample and for different return periods rp

1	boot_farr_fit.np(x, z, rp, B = 100)

`x`	the variable of interest in the counterfactual world.
`z`	the variable of interest in the factual world.
`rp`	the return periods for which the far is to be estimated.
`B`	the number of bootstrap samples to draw.

This function returns bootstrap non-parametric estimates of far, the fraction of attributable risk for records, as defined in Naveau et al (2018).The far is estimated from the bootstrap samples of the data x and z that are obtained by resampling bootstrap. The first bootstrap sample corresponds to the original dataset of x and z.

For the full reference, see : Naveau, P., Ribes, A., Zwiers, F., Hannart, A., Tuel, A., & Yiou, P. Revising return periods for record events in a climate event attribution context. J. Clim., 2018., https://doi.org/10.1175/JCLI-D-16-0752.1

An object of class ("boot_farrfit.np", "boot_farrfit", "farrfit"). It is a matrix where each columm contains the far estimated for the returns periods rp for a given bootstrap sample.

 library(evd)

 muF <-  1; xiF <- .15; sigmaF <-  1.412538 #  cst^(-xiF) # .05^(-xi1);
 # asymptotic limit for the far in this case with a Frechet distributiom
 boundFrechet <- 1 - sigmaF^(-1/xiF)
 # sample size
 size <- 100
 # level=.9
 set.seed(4)
 z = rgev(size, loc = (sigmaF), scale = xiF * sigmaF, shape = xiF)
 x = rgev(length(z), loc=(1), scale = xiF, shape=xiF)

 rp = seq(from = 2, to = 30, length = 200)

 # Resampling bootstrap for the non-parametrc estimation of the far
 boot_farr.np <- boot_farr_fit.np(x = x, z = z, rp = rp, B = 10)
 print(boot_farr.np)
 confint(boot_farr.np)

 ylim <- range(boundFrechet, boot_farr.np)
 plot(boot_farr.np, ylim = ylim, main = "boot far non-parametric")
 # Theoretical far for in this case (Z = sigmaF * X  with X ~ Frechet)
 lines(rp, frechet_farr(r = rp, sigma = sigmaF, xi = xiF), col = "red", lty = 2)
 abline(h = boundFrechet, col = "red", lty = 2)