ocd detector returning 0 for all test statistics

[multimeric]

Following the examples in the user guide, I have come up with this code:

require('ocd')

# Takes a data frame, trains the detector on the burn-in set, then starts detecting
segment.ocd = function(data, burnin, alpha){
  det = ChangepointDetector(dim=ncol(data), method='ocd', beta=1, thresh=c(11.6, 179.5, 54.9))
  det = setStatus(det, 'estimating')
  for(i in 1:burnin){
    row = unlist(data[i,])
    det = getData(det, row)
  }

  detector = setStatus(det, 'monitoring')
  changepoints = list()
  for(i in (burnin+1):nrow(data)){
    row = unlist(data[i,])
    det = getData(det, row)
    chpt = status(detector)
    if (is.numeric(chpt)){
      changepoints[[length(changepoints)+1]] = chpt
    }
  }
  print(changepoints)
  print(statistics(det))
}

# Makes an p-dimensional data.frame that shifts after a certain point in time
get.changepoint = function(dims=2, pre.args=list(n=50, mean=0, sd=1), post.args=list(n=50, mean=2, sd=1)){
  columns = list()
  for (i in 1:dims){
    columns[[letters[i]]] = c(do.call(rnorm, pre.args), do.call(rnorm, post.args))
  }
  data.frame(columns)
}

chpt = get.changepoint(dims=10, post.args=list(n=50, mean=2, sd=1))
segment.ocd(chpt, burnin=20, alpha=0.1)

So, I create a dataset that has a mean shift from 0 to 2 at n=50. This should be fairly easy to detect, however the changepoints list that I obtain is always empty, and the test statistics vector contains only 0s. What am I doing wrong here?

[yudongchen88]

In the second part of your segment.ocd function, detector should be changed into det (2 places). Then, you will see that the change is declared at time 31 (i.e. 50+1-burnin).