blueice2hotsea | November 22, 2013 at 7:44 pm |
Yes, you have it.
Though it appears the reason for selecting as prime a duo (or triple) has been missed.
The problem with trendology when you don’t know if there is an underlying periodic pattern is to avoid accidentally producing one. For example, if you chose 12 and 48, for some reason, on a 5-year (60 months = 12 + 48) moving average, you can pretty much bet you will on enough data create a false oscillation.
This overlay is visible in http://www.woodfortrees.org/plot/hadcrut4gl/mean:96/mean:24/mean:12 for example; note the artificially constructed regular sawtooth pattern on sections of the curve.
So a difference of 2 in 191 is, indeed, significant.