A panel of six change detectors on three more real series: one hit, one miss

Six change-point settings run on Hurricane Maria deaths, Argentina inflation and coal disasters. A spike is found by all six; a stabilisation is missed.

The free change-point detector has a --consensus mode: it runs six settings and reports how many agree on each change. The fourth batch ran it on three more real series with a documented event. Output: ventures/changepoint-detector/eval/consensus-real.out.txt.

SeriesDocumented eventResult
Puerto Rico daily deaths, 1,201 daysHurricane Maria, 2017-09-20inside a group found by 6 of 6 (span 09-20 to 10-08)
Argentina log CPI change, 79 quarters1989 hyperinflation peak1989Q2 by 6 of 6
Argentina log CPI change1991 Convertibility Lawnot listed at 3 of 6 or more
British coal disasters, 112 yearsCoal Mines Regulation Act 1887 (recalled)1887 by 4 of 6

What it says: a sharp spike is found by every setting, a gradual stabilisation in steps is not, and agreement among settings is not a detection rate. Every setting also agreed on three Puerto Rico changes we did not check against an event, and the monthly US saving rate returned 27 listed changes, so on long series the middle tier is a long list. Eleven real series, no labelled set.

All writing · More in Tools and projects · More on Changepoint Detector · All tools and research