Hi I am trying to check the existence of change points in the mean and in the variance using R. My data is not normally distributed, so I want to use the cusum test. Also I want the significance level to be 0.05.

I am trying to use the package "changepoints", but for change points in the mean I couldn't find a way of choosing the significance level because it says the asymptotic penalty is not available with cusum!

For the variance it lets me choose the significance level using: single.var.css(x,penalty="Asymptotic",value=0.05,class=TRUE,param.estimates=TRUE) however it always detects a change point. Sometimes it looks like a real changepoint, but when the series doesn't seem to have one it just detects a change point in the end of the series.

Anyone knows a way of calculating change points in R using the cusum test with a significance level of 0.05?

Thanks for helping!

  • 1
    $\begingroup$ Since the problem is resolved, I've cleared this comment thread. $\endgroup$ – user88 Oct 3 '11 at 15:49
  • $\begingroup$ @mbq, but how the problem was resolved? Was the solution in the comment thread? $\endgroup$ – German Demidov Feb 10 '16 at 14:32
  • $\begingroup$ It was about cross-posting of the question; no answer was there. $\endgroup$ – user88 Feb 11 '16 at 12:13
  • $\begingroup$ If it's not normally distributed, the variance is probably not a good metric. Maybe the inter-quartile range? $\endgroup$ – Jonas Lindeløv Jan 15 at 10:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.