distLexBoot function

Bootstrapping uncertainty intervals for return periods

Calculates and plots bootstrap uncertainty intervals for plotLextreme.

distLexBoot(
  dlf,
  nbest = 3,
  selection = NULL,
  n = 100,
  prop = 0.8,
  conf.lev = 0.95,
  replace = FALSE,
  RPs = NULL,
  log = TRUE,
  progbars = TRUE,
  quiet = FALSE
)

Arguments

• dlf: dlf object, as returned by distLextreme
• nbest: Number of best fitted distribution functions in dlf for which bootstrapping is to be done. Overridden by selection. DEFAULT: 3
• selection: Character vector with distribution function names to be used. Suggested to keep this low. DEFAULT: NULL
• n: Number of subsamples to be processed (computing time increases extraordinarily). DEFAULT: 100
• prop: Proportion of sample to be used in each run. DEFAULT: 0.8
• conf.lev: Confidence level (Proportion of subsamples within 'confidence interval'). Quantiles extracted from this value are passed to quantileMean. DEFAULT: 0.95
• replace: Logical: replace in each sample? DEFAULT: FALSE
• RPs: Return Period vector, by default calculated internally based on value of log. DEFAULT: NULL
• log: RPs suitable for plot on a logarithmic axis? DEFAULT: TRUE
• progbars: Show progress bar for Monte Carlo simulation? DEFAULT: TRUE
• quiet: Logical: suppress messages? See distLquantile. DEFAULT: FALSE

Returns

invisible dlf object, see printL. Additional elements are: exBootCL (confidence level), exBootRPs (x values for plot) exBootSim (all simulation results) and exBootCI (aggregated into CI band). The last two are each a list with a matrix (return levels)

Details

Has not been thoroughly tested yet. Bootstrapping defaults can probably be improved.

Examples

data(annMax)
dlf <- distLextreme(annMax, selection=c("gum","gev","wak","nor"))
dlfB <- distLexBoot(dlf, nbest=4, conf.lev=0.5, n=10) # n low for quick example tests
plotLexBoot(dlfB)
plotLexBoot(dlfB, selection=c("nor","gev"))
plotLexBoot(dlfB, selection=c("gum","gev","wak","nor"), order=FALSE)

See Also

plotLexBoot, distLextreme

Author(s)

Berry Boessenkool, berry-b@gmx.de , Sept 2015 + Dec 2016