Comments on: Pitchers!

By: David Gillies

David Gillies — Mon, 14 Dec 2009 00:57:02 +0000

James, sorry; I wasn’t trying to be rude. But I think we’d both agree that lay people would be amazed at the general shoddiness of academic codesmithing. Engineers tend to be better than scientists, probably because they approach the field as an engineering discipline rather than just a problem-solving tool. There are examples of good practise in scientific computing (I fondly remember a summer spent validating a corner of the Swedish Lund University code used to simulate the LEP particle accelerator: specifically the method for calculating the ‘thrust’ of an interaction, which is effectively the sum of the absolute values of parton momenta – I wrote a technique completely orthogonal to the standard one and validated the model, for low multiplicity events anyway, i.e. fewer than five or six jets. Most of the code had been checked in this fashion. FORTRAN 77 on a VAX 8650 under VMS/DCL with a greenscreen VT100 terminal! Happy days. And I know for most people this is the compsci equivalent of glossolalia).

By: S. Weasel

S. Weasel — Sun, 13 Dec 2009 22:42:53 +0000

Nethicus did a version of this that’s a lot better than mine and is going viral

On an happier note, though, I got t-shirts and postcards and greeting cards. Send your favorite hippie a taunting postcard!

By: Climate Change In Context « ricketyclick

Climate Change In Context « ricketyclick — Sun, 13 Dec 2009 17:08:58 +0000

[...] favorite weasel-girl did an early version of this using WUWT’s original [...]

By: James

James — Sun, 13 Dec 2009 17:04:17 +0000

David, I was not trying to make excuses, just explain how sloppiness turns into policy.
I agree that it is insane to try to plan intrusive and expensive public policies on the basis of monte-carlo programs that can’t even seem to get the sign of the change right (I’m back in Wisconsin now, and spent an hour chipping global warming off the sidewalk). If you want to look at our monte-carlos, feel free: pythia, herwig, etc are all available for scrutiny. (I remember running across some dumbness in Isajet, so I won’t claim they’re bug-free). Quite a lot of money was risked on those models when we built the detectors, but at least the models more or less work: under conditions X and Y pythia predicts Z, and when you do the experiment you find something that looks a lot like Z; so we guess it will still work when you have X and Y+epsilon. We have people devoted to refining the programs and trying to find the limits to their validity.
Where are CRU’s programs? Data analysis can be hairy because you have to know a lot about the measurement (calibration, variability, how it correlates with other measurements, failure rate, etc) before you know what it means, but the predictions should be a lot easier to understand.
If they can do something as simple as predict the average temperature in western India for the next decade, then maybe we can start talking. About whether the “warming/cooling/neither” is going to be a bad thing, for instance.
FWIW, you might be gratified to know that the software frameworks for our experiments were written by pros, and tested and code reviewed. The analysis modules–um… we find most of the bugs when funny things happen during use.

By: David Gillies

David Gillies — Sat, 12 Dec 2009 20:04:26 +0000

1) I assume anyone as screamingly funny as David Burge is by default of exceptionally high intelligence and probably polymathic. In my experience the smarter someone is, the funnier he is. The converse is true.

2) we don’t know. All we have is the time series analysis code and data from CRU. The GCMs are a mystery. They could be written with the same shocking level of incompetence as the CRU stuff; they could be shining examples of good software engineering practise. Unless and until we see the code we don’t now. GIGO is always there in potentia, but we can’t tell if it applies in this situation. We do know that the GCMs have failed to predict the decade long period we are in where there has been no net warming, but we cannot meaningfully critique them without the code and (almost as crucially) the boundary conditions.

By: Mrs. Peel

Mrs. Peel — Sat, 12 Dec 2009 19:16:50 +0000

heh, Iowahawk now has three links to his post in this thread alone. Time to move him up to the front page, Weas?

By: JD Will

JD Will — Sat, 12 Dec 2009 16:15:17 +0000

Davie Gilles,

I am also a software engineer, though my work (mfg/commercial) is not so grand as yours. I too was shocked at Harry’s travails, though I have cleaned up similar messes taking on some legacy systems.

I am interested in your opinion on two things:

1. This reconstruction of the reconstruction. Who knew IowaHawk was more than a hot rodding comic genius?
http://iowahawk.typepad.com/iowahawk/2009/12/fables-of-the-reconstruction.html#more

2. Wouldn’t the GIGO principle invalidate any analysis of the GCMs themselves? Harry’s read me told me that the data provenance was uncertifiable and cherry pickers could have had a field day with the ‘value added data’.
There also is lively web discussion going on as to the shifting nature of the land based temperature record.

It may truly be a house of cards, but I am waiting for the science to get better, because there is a basic prima facie argument for incremental AGW via GHG that has not been disproved .

Regards,

PS – Weasel, loved the GIF, spot on!

By: Waterhouse

Waterhouse — Sat, 12 Dec 2009 05:16:15 +0000

Now if only Zazzle made an electronic shirt that could display an animated .gif.

By: EW1(SG)

EW1(SG) — Sat, 12 Dec 2009 03:50:38 +0000

Mrs. Peel, you have my undying love and gratitude forever for leading me to this particular Burgeian nugget:

I'm also a big believer in learning by doing; if you really want to know how a carburetor works, nothing beats taking one apart and rebuilding it. That same rule applies to climate models.

I will treasure that amongst my stolen pet sayings for at least two weeks! /Oh! Shiny! Buh-bye!

By: Mrs. Peel

Mrs. Peel — Sat, 12 Dec 2009 02:53:45 +0000

Very educational post by Iowahawk, if you haven't seen it yet.