Global Warming Update: Phoenix (AZ) cancels flights as it's over 118 degrees for American Airlines regional planes

I'm not sure where this notion of 90% significance comes in, if you'd like to point me to a source. We tend to follow the p < 0.05 rule that most every field follows, which is itself arbitrary. There's lots of folks involved in bayesian statistics and signal-to-noise analysis in the field. Fisherian statistics at p < 0.1 (if that's what you're implying) is quite rare.

I agree 95% is arbitrary (and kind of low, considering how many similar tests can be made), but IPCC used 90% as their confidence terminology.

Confidence Terminology (Degree of confidence in being correct )

Very high confidence means: At least 9 out of 10 chance

IPCC is mostly noise wrote:

I'm not sure where this notion of 90% significance comes in, if you'd like to point me to a source. We tend to follow the p < 0.05 rule that most every field follows, which is itself arbitrary. There's lots of folks involved in bayesian statistics and signal-to-noise analysis in the field. Fisherian statistics at p < 0.1 (if that's what you're implying) is quite rare.

I agree 95% is arbitrary (and kind of low, considering how many similar tests can be made), but IPCC used 90% as their confidence terminology.

Confidence Terminology (Degree of confidence in being correct )

Very high confidence means: At least 9 out of 10 chance

http://www.ipcc.ch/publications_and_data/ar4/wg1/en/ch1s1-6.html

Thanks! I see what the original poster (or you) was getting at now. IPCC went with discrete qualitative ideas like "very likely" to keep things simpler, but a lot of the quantitative details get lost in that process.

I would argue that, from a probabilistic standpoint, 90% chance of something happening is a pretty solid threshold for "very likely."

This concept is unrelated to the p-value stuff I was getting at. The p-value discussion is really about statistical significance. This is more about a distribution of forecasts that fall within/above some range.

And won't admit he was wrong wrote:

Re: Inter-disciplinary work

If only Michael Mann had known anything about statistics....

I'm sure Michael Mann knows some things about statistics. That being said, anyone I know that has worked with or under him says he's a jerk.

Similar to any field/philosophy/religious group/political party, one annoyingly loud person doesn't represent the whole community.

General circulation models (climate models) with perturbations are a way to approximate the magnitude of natural variability and nail down the response of the climate system to a CO2 emissions experiment (which we have underway in the real world right now).

I suspect perturbations are the source of your conniptions.

As my second semester calculus prof noted decades ago, what good is an "approximation", if you have no bound on the error? None of these "perturbed" models came close to being reliable predictors in the last assessment (all 73 wrong).

Please indulge yourself by reading a physical oceanography or atmospheric dynamics publication. They are packed with advanced mathematics, since we do study a massive coupled fluid system

OK, I looked at Chapter 15 (seemingly the most relevant) in this textbook (#1 hit on search engine):

Largely this chapter is a very facial quasi-introduction to modelling (compare a math textbook if you like). The author does give some caveats about taking the models too seriously. E.g. "Calculations of turbulence are difficult" ... could just say that you rapidly (i.e. exponentially in the sense of chaos) approach noise and nonsense. The author does note the dimensionality problem spatially, but seems to under-rate it in the time evolution domain. The author then just points to some black-bo software, and it's hit-or-miss as to whether these are correct (and used as such) for your application. I think there is a total of one equation (labelled 15.1 in 15.4) in the entire section, which doesn't give me much confidence.

Admittedly this textbook is likely out of date on what can be simulated, but it's bold claim (evenmoreso a decade ago) that climate models can "simulate hundreds to thousands of years" is simply rubbish.

OK, maybe that's a fairly low-level textbook. So I did another web search, and came up with this "Advances in Computational Oceanography" paper (from 2006, but I don't think the datedness matters).

This looks more like a 10-page informational brochure than a serious paper (is it?). Again I don't see any actual math equations floating around, though occasionally some jargon pervades the text. The question of whether all these FVCOM modules actually are useful is not addressed (but having more Swiss army options is always a plus).

I'm not sure where this notion of 90% significance comes in,

Another example (IPCC definition)

“Extreme weather event. An extreme weather event is an event that is rare at a particular place and time of year. Definitions of rare vary, but an extreme weather event would normally be as rare as or rarer than the 10th or 90th percentile of a probability density function estimated from observations ... ”

The notion of "extreme weather event" as 20% of the distribution is probably not be what the typical listener would expect (especially as you'd likely need to make a Bonferroni type adjustment given the number of comparisons that could be made).

1/5 of the time wrote:

I'm not sure where this notion of 90% significance comes in,

Another example (IPCC definition)

“Extreme weather event. An extreme weather event is an event that is rare at a particular place and time of year. Definitions of rare vary, but an extreme weather event would normally be as rare as or rarer than the 10th or 90th percentile of a probability density function estimated from observations ... ”

Note too: convenient conflation of "extreme" with the cognate "extremely", which means 95% confidence at other junctures (or ~97%, given their example of rolling a dice [sic] twice and getting a six both times).