THE BOOK--Playing The Percentages In Baseball

Tuesday, August 24, 2010

Weather research paper

By Tangotiger, 03:35 PM

I just got this email, so I haven’t had a chance to read it:

Don’t know if you’ve seen it, but Blakeley McShane has a new paper on statistical analysis of climate research coming out that is making some waves on the net (search for McShane and Wyner on Google blog search to see some of the reaction). You can download it here: http://www.imstat.org/aoas/next_issue.html (direct link: http://www.e-publications.org/ims/submission/index.php/AOAS/user/submissionFile/6695?confirm=63ebfddf). McShane has collaborated with Shane Jensen in the past on some Bayesian analyses of baseball topics: http://www.blakemcshane.com/research. The new paper is very readable and interesting.

• Sabermetrics • Parks

#1 Dickson (see all posts) 2010/08/25 (Wed) @ 12:00

new paper????

#2 (see all posts) 2010/08/25 (Wed) @ 20:53

If anyone on this site ended up reading the paper and could briefly summarize it for me, I’d be grateful. What were the conclusions?

#3 Watercott (see all posts) 2010/08/27 (Fri) @ 15:44

Summary: Standard paleoclimatology research underestimates the uncertainty of their models. Evidence for this is that, among other things, semi-random data can be used to match the temperature record just as well as historical proxies (tree rings, bore holes, etc.) - there is a ton of advanced statistics going on too, about time and space correlation of the proxy data, problems with over-fitting, etc.

The conclusion is that a better way to model is with some Bayesian methods, and this leads to a historical model with extremely wide uncertainty bands. The ultimate conclusion is that there is only about an 80% chance that the last decade is the warmest in 1000 years, and a 36% chance that the last 30 years are the warmest.

Page 1 of 1 pages

<< Back to main

THE BOOK--Playing The Percentages In Baseball

Tuesday, August 24, 2010

Weather research paper

Latest...