R/statboot.R

In Jbrich95/sdfExtreme: Spatial Deformation for Non-Stationary Extremal Dependence

#' Stationary Bootstrap
#'
#' Creates one bootstrap sample of a \eqn{N} by \eqn{d} matrix, \code{X}. Here \eqn{N} denotes the number of time points
#' and \eqn{d} the number of sampling locations. The stationary bootstrap (Politis and Romano, 1994) generates a bootstrap sample
#' by repeated sampling of random blocks with expected value \code{block.size} until a sample of length \eqn{N} has been created.
#'
#' @param X Matrix with \eqn{N} rows corresponding to time points and \eqn{d} columns corresponding to spatial locations.
#' @param mean.block.size Expected value of temporal block size.
#' 
# @return
#'
#'@references Politis and Romano (1994), JASA,5(4):303-336,
#' (\href{https://doi.org/10.1080/01621459.1994.10476870}{doi})
#'
#'
#' @examples \eqn{N} by \eqn{d} matrix
#'
#' data(Aus_Heat)
#' Z<-Aus_Heat$Temp.
#'
#' unif<-function(x) rank(x)/(length(x)+1)
#' Z_U<-Z
#' for(i in 1:dim(Z_U)[2]) Z_U[,i]<-unif(Z[,i]) # Transform to uniform margins
#'
#' q<-0.95
#'
#' ind.triple<-c(1,2,3) ##Denotes indices of triple for which triple-wise chi calculated.
#'
#' block.mean<-14 #Mean block size for random block choice - Here a fortnight
#'
#'boot <- stat.boot(Z_U[,ind.triple],block.mean)
#'
#'#Create one bootstrap estimate of triple-wise chi
#'chi3.emp(boot[,1],boot[,2],boot[,3],q)
#'

stat.boot=function(X,mean.block.size){
  N=dim(X)[1]
  
  #Generate random block sizes
  b=rgeom(N,1/(mean.block.size))+1
  
  sum=cumsum(b)
  
  b=b[1:min(which(sum>=N))]
  
  #Find starting indices
  inds=sample(1:N,length(b))
  all_inds=c()
  for(i in 1:length(b)){
    
    block_inds=inds[i]:(inds[i]+b[i]-1)
    
    #Wrap around indices
    if(sum(block_inds>N)>0){
      block_inds[block_inds>N]=1:sum(block_inds>N)
    }
    all_inds=c(all_inds,block_inds)
    
  }
  
  #Cut down to N only
  all_inds=all_inds[1:N]
  
  boot=X[all_inds,]
  return(boot)
}