modifiedRanks <- function(plateau, binaryLevel, inverse=FALSE, penalty=0){
  
  if( ! missing(binaryLevel)  ){
    
    ##cat(" using: binary approach with parameter:", binaryLevel, " and inverse:", inverse, "\n")
    return( function(x){
      if( plateau < 1 ){
        stop(paste(" invalid plateau:", plateau))
      }
      if( inverse ){
        tmp <- rank(x, ties.method="min")
      }else{
        tmp <- rank(-x, ties.method="min")
      }
      tmp[ tmp>plateau ] <- binaryLevel
      tmp[ tmp<=plateau ] <- 0
      return(tmp)
    })
    
   }else{
     
     ##cat(" using: rank approach with plateau:", plateau, " and penalty:", penalty, " and inverse:", inverse, "\n")
     return( function(x){
       if( plateau < 1 ){
         stop(paste(" invalid plateau:", plateau))
       }
       if( inverse ){
         tmp <- rank(x)
       }else{
         tmp <- rank(-x)
       }
       tmp[ tmp>plateau ] <- plateau+penalty
       return(tmp)
     })
    }
}


calculateStatistics <- function(rawExpr,windowSize,maxCluster,...){
  
  probenzahl=nrow(rawExpr)
  genzahl=ncol(rawExpr)                      ## differs for nullstat !
  geneNames <- colnames(rawExpr)

  if( missing(maxCluster) ){
    maxCluster=ceiling(probenzahl/2)
  }else{
    stopifnot( maxCluster <= ceiling(probenzahl/2) )
  }
  
  resultStatistiken <- array(data=NA,
                             dim=c(maxCluster,(genzahl-(windowSize-1)),2),
                             dimnames=list(NULL,
                               geneNames[1:(genzahl-(windowSize-1))],NULL))
  for(inverse in 1:2){
    for(cp in 1:maxCluster){
      
      myRank <- modifiedRanks(cp, inverse=(inverse==2), ...)
      expr <- apply(rawExpr,2,myRank)             ## samples x genes
      
      diffInit = apply(expr[,1:windowSize],1,sum)## init
      diff = cbind(diffInit,expr[,(windowSize+1):genzahl] -
        expr[,1:(genzahl-windowSize)]   )
      
      statistiken = t( apply(diff,1,cumsum) )  ## still samples x genes
      statistiken = apply(statistiken,2,sort)
      
      resultStatistiken[cp,,inverse]<-apply(statistiken[(1:cp),,drop=FALSE],2,sum)
      
    }
  }
  return(resultStatistiken)
  
}