boot.sequence {extRemes} | R Documentation |
Bootstrap a sequence of exceedances.
boot.sequence(tmat, ymat, u)
tmat |
Output from boot.matrix ('iet' component). |
ymat |
Output from boot.matrix ('ce' component. |
u |
Threshold above which the sequences of exceedances are to be bootstrapped. |
The bootstrapped sequence contains the same number of clusters as identified in the original data. Clusters of exceedances and inter-exceedance times within clusters are resampled as single entities; inter-exceedance times between clusters are resampled independently. The values of non-exceedances are set equal to the threshold `u'.
Vector of exceedances above 'u'.
The bootstrapped sequence can be longer than the original sequence.
Maintained by Eric Gilleland.
Chris Ferro
Ferro CAT and Segers J. (2003). Inference for clusters of extreme values. Journal of the Royal Statistical Society B 65:545–556.
# Simulate 1000 uniform random variables. x <- runif(1000) # Perform runs declustering with run length = 1 and 90th percentile as threshold. u <- quantile(x, 0.9) z <- x > u dec <- decluster.runs(z, 1) # Make sure the estimated run length is not zero before doing the rest. if( dec[["par"]] != 0) { # Set up the matrices for bootstrapping. mat <- boot.matrix(dec, x) # Bootstrap with 500 iterations. eib <- numeric(500) for( i in 1:500) { set.seed(i) zb <- boot.sequence(mat[[1]],mat[[2]],u) > u eib[i] <- exi.intervals(zb) } # end of for 'i' loop. # Obtain bootstrapped 95th percentile confidence intervals. conf.int <- quantile( eib, c((1-0.95)/2,(1+0.95)/2)) } # end of if run length not zero stmt.