CODEWITHSUNDEEP
rmarkov-chainsmarkov
Rup Mitra
Rup Mitra

Reputation: 41

Incorrect Number of probabilities simulating a Markov chain

my transition probability matrix is like this 

             BP          IP             SP 


BPBP     0.4586757     0.3772354     0.1640889

IPBP     0.3489484     0.4746654     0.1763862

SPBP     0.3756522     0.4162319     0.2081159

BPIP     0.3646061     0.4640000     0.1713939

IPIP     0.2666122     0.5654956     0.1678922

SPIP     0.3054187     0.4860427     0.2085386

BPSP     0.4125561     0.3974215     0.1900224

IPSP     0.2974337     0.5069415     0.1956247

SPSP     0.3576642     0.4333942     0.2089416

and the code to simulate a first order MC from it is

  function(trans,initprob,N)
  {
  BrokerPosition <- c("BP", "IP", "SP")
  mysequence<-character()
  firstposition <- sample(BrokerPosition, 1, rep=TRUE, prob=initprob)
  mysequence[1]   <- firstposition
  for (i in 2:N){
   prevposition <- mysequence[i-1]
   probabilities  <- trans[,prevposition]
   BPosition<- sample(BrokerPosition, 1, rep=TRUE, prob=probabilities)
   mysequence[i]  <- BPosition
                }
  return(mysequence)
  }

but since this is a non square matrix I am getting an error of mismatch of probabilities ,any idea how to solve this

Upvotes: 2

Views: 661

Answers (1)

Vincent Zoonekynd
Vincent Zoonekynd

Reputation: 32351

You are using your transition matrix in the wrong direction: try with 

trans[ paste( mysequence[c(i-2,i-1)], collapse="" ), ]

An alternative would be to to convert your second order Markov chain into a first order one: for instance, the states after IPBP would be BPIP, BPBP, BPSP (and the other ones, IP* and SP*, would have zero probabilities). The transition matrix is then a 9*9 matrix with a lot of zeroes.

Upvotes: 2

Related Questions