How can this series pass the structural change test?

I have this series:

r <- c(14.1649994,17.3997831,32.8877030,6.1547551,6.3594748,101.9584077,196.8411517,23.8292043,11.7583992,2.2915478,5.5088687,16.8188390,2.3077979,2.6666377,2.0955362,0.9319688,0.6688078,1.0523184,6.8158264,13.1813211,17.8467944,1.9876636,7.0219306,3.9272217,4.3325448,7.8778125,1.1878058,1.7207762,2.9655665,5.8571428)


I'm using the strucchange package to detect the structural changes. The result is:

> sctest(r~1)

Recursive CUSUM test

data:  r ~ 1
S = 0.7785, p-value = 0.1579


How is it possible that the test didn't detect changes? There is a very big move rising around 200.

Take a look at the plot of the series below.

-
sctest(r~1, type="Chow", point=7) – Brandon Bertelsen Sep 6 '11 at 21:45
@Brandon, I have to do it programmatically, I don't know the point with the break, in this case yes because I have checked it visually, but in the case I can't see it, what method should i use? how is it possible this break is not detected? – Dail Sep 6 '11 at 21:52
To me, an outlier test (like Grubbs) seems more approriate. Run this code and take a peak to see why your test "seems" to fail. r.cus <- efp(r~1, type="Rec-CUSUM") and then plot(r.cus). Here's a related paper on the topic. blogperso.univ-rennes1.fr/arthur.charpentier/public/… – Brandon Bertelsen Sep 6 '11 at 22:19
@Brandon I have seen that function but seem there is not a p-value to understand if the test pass or not. – Dail Sep 7 '11 at 6:36
library(outliers);grubbs.test(r); grubbs.test(r)\$p-value;dixon.test(r,type=22) note that these only test to see if there is ONE outlier. – Brandon Bertelsen Sep 7 '11 at 7:11