tag:blogger.com,1999:blog-7729093380675162051.post5734284638721969154..comments2024-03-28T13:56:47.604+11:00Comments on moyhu: Uncertainty of temperature trends.Nick Stokeshttp://www.blogger.com/profile/06377413236983002873noreply@blogger.comBlogger4125tag:blogger.com,1999:blog-7729093380675162051.post-29795392067336987542013-11-04T13:11:21.147+11:002013-11-04T13:11:21.147+11:00Thanks, Jeff.
I played around with the various ap...Thanks, Jeff. <br />I played around with the various approximations <a href="http://www.moyhu.blogspot.com.au/2013/09/adjusting-temperature-series-stats-for.html" rel="nofollow">here</a>. I've been hatching a post which says more strongly that if you identify the smooth (non-AR) part of the deviation, there isn't much ARIMA left. That's not good news, really, because there's still uncertainty - just harder to quantify.<br /><br />Nick Stokeshttps://www.blogger.com/profile/06377413236983002873noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-50781612783860857562013-11-04T11:24:42.600+11:002013-11-04T11:24:42.600+11:00Nick,
This is a fantastic post. Your explanatio...Nick, <br /><br />This is a fantastic post. Your explanations of the Toplitz matrices were very clear. I love simplification of something otherwise complex. <br /><br />I see different AR models as vague approximations of physical series. It is hard to associate too much credence to ANY of the results regarding significance, except for the fact that if ANY are close to significant, it often is reasonable. Your point at WUWT regarding monthly temperature series needing an extra chunk of variance taken from them is understandable. All of the temperature series suffer from more than a bit of change in recent annual anomaly in their signal. I'm not sure what the ultimate advantage is either way.<br />Jeffhttps://www.blogger.com/profile/00102232063298547788noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-8981263127502239902013-09-05T21:30:55.318+10:002013-09-05T21:30:55.318+10:00I've tried just curve fitting trend removal - ...I've tried just curve fitting trend removal - using Fourier peaks. The idea is to explain some of the variation by fitted curves rather than stochastic models, and reduce the oscillations in the acf. But yes, the simplest next step is probably to do the same analysis on the F&R adjusted series.Nick Stokeshttps://www.blogger.com/profile/06377413236983002873noreply@blogger.comtag:blogger.com,1999:blog-7729093380675162051.post-86683425200581024932013-09-05T21:13:04.723+10:002013-09-05T21:13:04.723+10:00> There is a marked periodicity at about 45 mon...> There is a marked periodicity at about 45 months<br /><br />ENSO ish? What if you try with the F+R ENSO removal method?William M. Connolleyhttps://www.blogger.com/profile/05836299130680534926noreply@blogger.com