检索绘制的变更点的坐标

Question

I have plotted Change Points on R as

values <- c(1, 2, 3, 4, 5, 6, 7, 8, 3, 1, 2, 3 )
values.ts = ts(values, frequency = 12, start = c(2017, 1))
chpoints = cpt.mean(values.ts, method="PELT")
cpts(chpoints)
plot(chpoints)

and the plotted graph is

The cpts(chpoints) returns [1] 4 8 .

Question 1: What are [1] 4 8 ?

Question 2: How can i get the coordinates of plotted change points, x,y coordinates of the red lines?

Answer 1

Quesion 1: The cpts(chpoints) returns vector of segments' end positions. Following your example:

plot(chpoints)
abline(v=index(values.ts)[cpts(chpoints)[1]], col="blue") # 4
abline(v=index(values.ts)[cpts(chpoints)[2]], col="green") # 8

will add two vertical lines which indicate segment with 'stable' mean. There are no. of segments - 1 change points ( [1] 4 8 ) as the last 'change point' is the last value of your time series.

Question 2: Date coordinate for i-th break can be found as:

index(values.ts)[cpts(chpoints)[i]]

'Stable' mean values are stored here:

param.est(chpoints) # 2.50 6.50 2.25

I still don't understand why there are white spaces after change points (red lines do not cover the whole time-series). However, if we compute mean values manually:

mean(values.ts[1:cpts(chpoints)[1]] ) #
mean(values.ts[(cpts(chpoints)[1]+1):cpts(chpoints)[2]] )
mean(values.ts[(cpts(chpoints)[2]+1):length(values.ts)] ) 

we receive 2.50 6.50 2.25 as above.