Comments on: Signal to Noise Ratio Estimates of Mann08 Temperature Proxy Data

By: A repy to Dr. Jim Bouldin « the Air Vent

A repy to Dr. Jim Bouldin « the Air Vent — Thu, 28 Jul 2011 03:14:23 +0000

[...] that would pass if you accept Mann’s incredibly generous autocorrelation assumptions. Far, far, higher if you do not. One of the most difficult scams of the paper is the determination of the correct autocorrelations to [...]

By: Jeff Id

Jeff Id — Wed, 25 Aug 2010 22:07:32 +0000

How about moisture vs temp.

By: Willis Eschenbach

Willis Eschenbach — Wed, 25 Aug 2010 19:15:06 +0000

Teleconnections? …

By: Jeff Id

Jeff Id — Sat, 21 Aug 2010 02:02:29 +0000

`Willis,

I hope this properly replies to your comment. I’ve had trouble lately. I’d forgotten about your method, I’ve got better chops now for these posts then I did two years ago. It was really quite brilliant, maybe your best.

Think about this, considering that you came up with 1:4 (maybe 1:5) in ‘common’ signal of messy data, it’s interesting that a different method comes up with 1/12 in temperature signal. It makes me wonder what the actual 1/4 signal might be comprised of.

By: Willis Eschenbach

Willis Eschenbach — Sat, 21 Aug 2010 01:08:36 +0000

Jeff, I don't know if you've seen my post on the topic of the signal to noise in the Mann 2008 proxies. I found the common signal / noise ratio to be about 1 – 4 (one part signal to four parts noise).

By: Scott Brim

Scott Brim — Fri, 20 Aug 2010 16:06:41 +0000

Jeff Id (Aug 19 17:45), Re: Jeff Id (Aug 19 17:45), OK, that's a start. You have the dimensional units and a few caveats listed. But recognize as well that your definition of the term "signal" might be specific and unique -- to some greater or lesser degree -- to the context of this particular analysis. Therefore we need additional detail concerning the physical phenomena the term "signal" is intended to characterize; the underlying theoretical basis which allows its employment within this particular analysis; and the term's historical provenance as used in previous scientific literature. If your work is to be fully auditable from a nuclear-grade QA perspective, then the information needs to be provided in a more detailed and precise form.

By: EdeF

EdeF — Fri, 20 Aug 2010 15:21:11 +0000

Jeff, in the real world you need a signal-to-noise ratio of about 12-1 for
good reception of a communications channel for instance. Any degradation from that means dropped words, bits, etc. The global positioning system signal is
very faint, but is pulled out of the noise because it is coded with a long
pseudo-random code that when matched on reception boosts the signal by 70db.
We don’t know a priori what the “climate signal” looks like. We can’t boost it out of the noise with fancy processing. Correlations fo 0.1 and negative S/N
ratios mean we have noise in noise, pseudocorrelations if you will.

By: Steven Mosher

Steven Mosher — Fri, 20 Aug 2010 05:31:58 +0000

once you figure out apply() and aggregate() you will never loop again.

By: Benjamin

Benjamin — Fri, 20 Aug 2010 03:15:13 +0000

Why don’t consider that proxies can act like low-pass filter for the temperature signal ?

High frequency is just some noise.
Low frequency is the temperature signal but the 100yrs of the instrumental temperature record is just not long enough to extract it (when calibrating).

So basically what people are doing when selecting their proxies is trying to get good agreement between noise and instrumental temperature… and then extrapolate that back to medieval time.

By: Jeff Id

Jeff Id — Fri, 20 Aug 2010 02:15:36 +0000

Here is a link to the plot you requested. It took hours to run, but was interesting.

http://noconsensus.files.wordpress.com/2010/08/proxy-snr.jpg