findCP: Wisely Pick Changepoints
In onlineCPD: Detect Changepoints in Multivariate Time Series

Description Usage Arguments Details Value Author(s) See Also Examples

Function that helps narrow down possible changepoints. Sometimes the onlineCPD and offlineCPD functions report an excessive number of changepoints. This function is called inside plot.oCPD and removes them based on criteria described in Details.

1	findCP(oCPD, buffer = 10)

`oCPD`	results of a call to `onlineCPD` or `offlineCPD`, this is the time series that the function will choose changepoints for.
`buffer`	a whole number. Any time there are two changepoints within `buffer` data points, `findCP` will choose one to replace them both.

The function chooses changepoints based on the following criteria; first, if there are two changepoints within buffer data points, it will pick one to replace them both, based on the sum of the probabilities for each change. Then, if the algorithm ever changes it's mind on a changepoint (i.e. the red, diagonal line ever jumps up) the function will remove the changepoint.

A vector of changepoints.

Zachary Zanussi

plot.oCPD

## Two changepoints within \code{buffer} points:
set.seed(5)
y <- c(rnorm(15,0.5,0.01),0.505,0.51,0.015,0.52,rnorm(15,0.6,0.05))
res <- offlineCPD(y)
plot(res,buffer=10)     ## Will replace the two changepoints
plot(res,cleanCP=FALSE) ## Will not replace the changepoints

## A changepoint is forgotten:
set.seed(5)
y <- c(rnorm(50,0.5,0.1),rnorm(20,0.48,0.02),rnorm(50,0.5,0.1),rnorm(75,0.7,0.4))
res <- offlineCPD(y)
plot(res)               ## Will remove the changepoint that has been second-guessed
plot(res,cleanCP=FALSE) ## Will not remove the changepoint