Is anyone clean?
Looking at some of the e-mails in my inbox, I'm pretty sure someone could pull a Climategate on them if they were leaked. "Oh riiight, you were just 'cleaning the data' to 'eliminate' the 'drop-out effect.'" Headlines next day: "How psychology is built on one MASSIVE lie!" "Researchers scrub data to keep grant money flowing." Nebuchadnezzar (talk) 23:59, 30 June 2011 (UTC)
"The CRU data was also independently replicated." I can't see anything in the references to justify this statement. Can anyone point me to evidence that there has been independent replication of the temperature reconstructions reproduced by the Hadley CRU / NASA GISS Group / University of Arizona teams?— Unsigned, by: 18.104.22.168 / talk / contribs 01:27, 1 June 2014 (UTC)
- The Muir Russell inquiry (which was comparatively negative in regards to how the scientists "display[ed] the proper degree of openness") came to the same conclusions using GHCN and NCAR data. Osaka Sun (talk) 02:48, 1 June 2014 (UTC)
If you have a piece of software with bugs in it that causes it to manufacture a warming trend, it doesn't matter how many different sets of data you run through it, you will always see a warming trend. The Hadley CRU / NASA GISS Group / University of Arizona teams are all using the same code library to do their reconstructions; ergo, it doesn't matter what data they run through it, it's always going to show warming. Independently replicated means that someone unconnected with them has written their own code and with it replicated their results. Can you point me to someone who has done this? I know of two cases where attempted replication failed; I know of none where it has succeeded. — Unsigned, by: 22.214.171.124 / talk / contribs 04:38, 1 June 2014 (UTC)
I have a Bachelor of Mathematics and 35 years experience as a programmer. I have a fair idea of what is involved in climate modelling. I have worked at a university and seen academics construct theories based on faulty software that have collapsed when the bugs have been found. And I'm not interested in arguing with crack pots about their propaganda. I'm asking if anyone can point me to a paper that demonstrates independent replication of the temperature reconstruction.— Unsigned, by: 126.96.36.199 / talk / contribs 09:15, 1 June 2014 (UTC)
- Here's a couple, including the Russell inquiry. Have fun.
- Interestingly, "point me to a paper that demonstrates independent replication of the temperature reconstruction" sounds a lot like like this argument. Osaka Sun (talk) 10:06, 1 June 2014 (UTC)
Re: 11 second pause. Yes, that sort of thing is bullshit. :-) What I'm looking for is replication, ie a paper where someone took the surface temperature record, ran it through their own separately written software, and got the same result. I haven't seen one yet.
-  for the claims that a) all three listed groups use the same piece of software in all their reconstructions, and b) the software in question has a warming bias.--ZooGuard (talk) 10:17, 1 June 2014 (UTC)
(a) "use common code"; read the climategate emails to see that this is the case. They talk about sharing newly written procedures between the various groups. That's not the same as "use the same program," a claim that would obviously be nonsense. (b) I don't know whether their software introduces a warming bias; that's why I want to see an independent replication. The independent reconstructions I have seen don't replicate their results.
- Riddle-me-this, though, since it's important. What if someone was to hypothetically replicate it and show a "wrong" trend? What then? Do you shout "aha! This global warming malarky is a scam!"? Because that wouldn't actually prove anything, it could well be that the "independent replication" was the flawed code. If you're looking for errors, you need to be looking at errors in the model, not the code, and looking at errors in any underlying assumptions (the total limit for oceans acting as a carbon sink, for instance, is very difficult to model). Don't click here 13:22, 1 June 2014 (UTC)
If someone fails to replicate that indicates a problem that needs to be addressed by further attempts to replicate. Replicability is a requirement of evidence; it has to be demonstrated, not just assumed.
I had a look at some of the papers in the list given above (I won't waste my time with all of them) and found what I expected. One of the standard signs of a crackpot theory is the gathering of a particular type of evidence: if the theory is true, X will be happening, so showing that X is happening proves that the theory is true. In this case: (a) assume the theory is true and that the initial reconstruction is correct; (b) find some natural phenomenon that correlates with the initial reconstruction and declare it to be a temperature surrogate; (c) produce a graph of that phenomenon and claim that you have replicated the initial reconstruction. Carefully ignore: (1) correlation is not causation; and (2) the claim in built on the assumption that the theory is true. If the theory is false and the initial reconstruction is wrong, then there is no correlation and the rest of it falls apart. So this is not even supporting evidence, let alone replication.
I can't see anything in this list that actually replicates the initial reconstruction from the instrumental record. Can someone point me to one? Or is it actually the case that no such reconstruction exists?
- Why should we do this for you, exactly? If such a reconstruction existed, would that change your opinion about the fact of global Climate Change? Hipocrite (talk) 14:42, 2 June 2014 (UTC)
- PS: . Hipocrite (talk) 14:44, 2 June 2014 (UTC)
Missing the main point
Attempts to dismiss the entire Climategate scandal as a hoax by way of dismissing individual points such as the 'hide the decline' or 'trick' quotes overlook the elephant in the room, which is that the overall tone of the emails suggests work of a substandard quality. This matters far more than any one example of supposed misconduct.
Policy changes costing trillions of dollars were enacted on the strength of CRU and similar climate reports, on the understanding that these reports constituted science of the best possible quality. Yet, Climategate showed that research in a rather different light, as being fragmentary data full of holes and discrepancies that had been patched and pieced together, missing parts being extrapolated, and obviously faulty sections snipped out. When seen in that light, you have to ask whether policies as expensive as those should be based on data of such mediocre integrity. --Anteaus (talk) 16:01, 20 July 2014 (UTC)
- Science is judged by the quality of its methodology, not by the "tone" set by researchers in emails. The only evidence that I see is that standard statistical and mathematical procedures were used to account for known inaccurate and faulty data, and in fact this strengthened the arguments made by the researchers. The fact that the researchers' methodology is sound is all that really matters, so I don't follow how you can claim their research was substandard based on the tone of a few emails. - Grant (talk) 16:09, 20 July 2014 (UTC)
Trick, treat, or statistical numptiness?
"if a data series ends in 1980, then 4-5 years of model data,...... needs to be incorporated to get a reliable moving average to end in 1980, rather than 5 years previous" The weasel word here is 'reliable'. Yes, you can add the extra data from a model. No, you can't describe this as reliable. If you want 5 years' worth of a moving average, your reliable data will be 5 years old. Doublenegative (talk) 08:03, 10 August 2014 (UTC)
Trick or stick?
Ok, so the "trick" was no fraud, but it still looks shaky methodology (assuming that it is explained correctly here). Extrapolation of experimental data is always to be handled with caution. Adding extra data points to your data set that are not homogeneous with it, and extrapolating on that, is asking for trouble.
That holds true even if the added data are "better" than the original data set. If the original data set has a systematic error (which you say is indeed the case here!) and the added data points do not have the same systematic error, you're definitely introducing artifacts into the graph.
A simple example to demonstrate that: let's say I have a series of temperatures deduced from a proxy, which is constant at 20°. Then I have a series of temperatures measured directly, which is constant at 21° -- that is, the proxy series has a systematic error of -1°. So I augment the proxy series by adding at the end four data points from the real temperatures series -- because those are "better" data, so it's ok to use them, isn't it?
In fact, it isn't. The augmented series runs like this:
Now compute the 5-years running average:
Voila: I've created a "hockey stick" graph! (Does that ring a bell?)