R/cyclical.min.max.R

In ORIClust: Order-Restricted Information Criterion-Based Clustering Algorithm

cyclical.min.max <-
function(data,x,n.rep,min1,max1){ 

   k <- length(n.rep)
   data <- as.numeric(data)
   n <- sum(n.rep)

   xa <- x
   waa <- n.rep
   xaa <- xa*waa
  
   xpb <- rep(0,k)

   h1 <- isodecre(xaa[1:(min1-1)],waa[1:(min1-1)])

   if((min1+1)<=(max1-1))
      h2 <- isoincre(xaa[(min1+1):(max1-1)],waa[(min1+1):(max1-1)])
   if((min1+1)>(max1-1))
      h2 <- NULL


   h3 <- isodecre(xaa[(max1+1):k],waa[(max1+1):k])

   peak1 <- min(c(xa[min1],h1[length(h1)],h2[1]))
   peak2 <- max(c(xa[max1],h2[length(h2)],h3[1]))
   xpb <- c(h1,peak1,h2,peak2,h3)

   is <- xpb
   sig <- 0;
   len1 <- 1;
   len2 <- 0;
   for(i in 1:k){
      muu <- is[i]
      len2 <- len2+n.rep[i]
      x1 <- data[len1:len2]
      len1 <- len1+n.rep[i]
      sig <- sig+sum((x1-muu)^2)
   }
   sig <- sig/n;

 like <- -0.5*n*log(2*pi)-0.5*n*log(sig)-0.5*n

list( logelr=like, mu=is)
}