[R] fine tune changepoint

[Luigi Marongiu]

Dear all,
I am using the package changepoint to identify a cut-off between positive
and negative values for an assay. I followed the vignette of the package
and the results are replicated, so I tried with a dataset of my choice. In
the assay, the positive have a value above 0.8 and the negative below 0.6,
so i did:

set.seed(10)

my.data <- c(rnorm(100, 0.8, 0.1), rnorm(100, 0.6, 0.1))

plot(my.data)


cp <- cpt.mean(my.data, method='PELT')

cpts(cp)

plot(cp, type='l', cpt.col='blue', xlab="Index", cpt.width=4)

It is visible that the data is clearly separated in two clusters, but the
algorithm did not find any change point. I tried also the other methods of
the function but no change point was found.

How can I fine tune the analysis to identify a change point to use as
cut-off? such fine-tuning should also be generic because should be applied
to any dataset.

thank you

L.

	[[alternative HTML version deleted]]