WBS.univar {changepoints} | R Documentation |
Perform wild binary segmentation for univariate mean change points detection.
WBS.univar(y, s, e, Alpha, Beta, delta = 2, level = 0)
y |
A |
s |
A |
e |
A |
Alpha |
A |
Beta |
A |
delta |
A positive |
level |
Should be fixed as 0. |
An object of class
"BS", which is a list
with the following structure:
S |
A vector of estimated change point locations (sorted in strictly increasing order). |
Dval |
A vector of values of CUSUM statistic. |
Level |
A vector representing the levels at which each change point is detected. |
Parent |
A matrix with the starting indices on the first row and the ending indices on the second row. |
Haotian Xu
Wang, Yu and Rinaldo (2020) <doi:10.1214/20-EJS1710>.
thresholdBS
for obtaining change points estimation, tuneBSunivar
for a tuning version.
set.seed(0)
cpt_true = c(20, 50, 170)
y = rnorm(300) + c(rep(0,20),rep(2,30),rep(0,120),rep(2,130))
intervals = WBS.intervals(M = 300, lower = 1, upper = length(y))
temp = WBS.univar(y, 1, length(y), intervals$Alpha, intervals$Beta, delta = 5)
plot.ts(y)
points(x = tail(temp$S[order(temp$Dval)],4),
y = y[tail(temp$S[order(temp$Dval)],4)], col = "red")
WBS_result = thresholdBS(temp, tau = 4)
print(WBS_result$BS_tree, "value")
plot(WBS_result$BS_tree)
print(WBS_result$BS_tree_trimmed, "value")
plot(WBS_result$BS_tree_trimmed)
cpt_hat = sort(WBS_result$cpt_hat[,1]) # the threshold tau is specified to be 4
Hausdorff.dist(cpt_hat, cpt_true)
cpt_LR = local.refine.univar(cpt_hat, y)
Hausdorff.dist(cpt_LR, cpt_true)