mt.hmm<-function(lfdr, q)
{

## USAGE
 # mt.hmm(lfdr, q)
 
## ARGUMENTS
 # lfdr: local false discovery rate sequence
 # q: the FDR level

## DETAILS
 # mt1.hmm gives a multiple testing rule in a hidden markov model
 # --that controls the FDR at level q, based on sequence of lfdr

## VALUES
 # nr: the number of rejected hypotheses
 # th: the threshold
 # re: the rejected hypotheses
 # ac: the accepted hypotheses
 # de: the decision rule

  m=length(lfdr)
  st.lfdr<-sort(lfdr)
  hps<-rep(0, m)
  if (min(lfdr)>q)
  {
    k<-0
    threshold<-1
    reject<-NULL
    accept<-1:m
  }
  else
  {
    k=1
    while(k<m && (1/k)*sum(st.lfdr[1:k])<q){
      k=k+1
    }
    k<-k-1
    threshold<-st.lfdr[k]
    reject<-which(lfdr<=threshold)
    accept<-which(lfdr>threshold)
    hps[reject]<-1
  }
  y<-list(nr=k, th=threshold, re=reject, ac=accept, de=hps)
  return (y)

}