When you throw more and more variables into a regression without theory to back them up in order to get a significant result, that is called a fishing expedition.

When (a la Mann) you also select which data series to allow into the estimation in addition to thowing in variables at will, that is called shooting fish in a barrel.

It is used to refer to the practice of looking at the data (graphing, computing descriptive statistics, running regressions, reading previous studies using the same data set etc) and *then* coming up with hypotheses that match the patterns in the data.

Since the hypotheses are formed to conform to whatever is seen in the data, there is no chance they can be rejected. Now, if the data set were used to form ideas about a model to be tested, then a separate but equally applicable data set were obtained and the model were tested against those data, then that would be a proper application.

If you torture a data set enough, it will confess to what you want to hear.

When I look at climate related work, I do not see many testable hypotheses. I see a lot of maintained hypotheses. And, an alarming tendency to equate correlation with causation.

The idea seems to be to search for ways to construct series that correlate with human activity without seriously questioning the connection between actual temperature movements and the constructed series.

Sinan

Guess I better go over and use the power of Wiki to fix that erroneous information.

“Data snooping” is another term which may be more apt for what’s going on with the multiproxy studies. The point is that Osborn and Briffa or Hegerl know in advance what particular proxies look like and so the data has been “snooped” in advance and just as in the econometric cases, statistical tests no longer apply as they would to fresh data.

The term Data-mining I am familiar with, is a method to scan corporate archives for useful legacy data. Wiki describes misuse of statistical data as “data-dredging” or “data-fishing”.

Data mining has been defined as “the nontrivial extraction of implicit, previously unknown, and potentially useful information from data” [1] and “the science of extracting useful information from large data sets or databases” [2].

Used in the technical context of data warehousing and analysis, the term “data mining” is neutral. However, it sometimes has a more pejorative usage that implies imposing patterns (and particularly causal relationships) on data where none exist.[citation needed] This imposition of irrelevant, misleading or trivial attribute correlation is more properly criticized as “data dredging” in the statistical literature. Another term for this misuse of statistics is data fishing.

http://en.wikipedia.org/wiki/Data_mining

So, it often doesn’t matter whether people know one another or not, or have met at conferences. It matters whether someone is advancing ideas that contradict one’s own, and what one thinks about the meaning of the relevant data. Ideas are, after all, the currency of science. Those who have the best ones are the richest. That’s what the fight’s all about, and the passions are real.

People of good personality, like Rob Wilson, almost always will gracefully concede if they’re shown wrong through an unarguable analysis. Others will show bad grace, right or wrong. The personal lesson from practicing science is that it is always better to not be ego-involved in the outcome. The challenge is to be thoughtful, not to be right. Nature decides who’s right, and nature is always trickier than anyone knows.

The critical ingredient leading to clarity, as always, is that data and methods be fully exposed. I’ve never, ever, seen anything in chemistry like the deliberate and dishonest obscurantism that infects climate science.

Wouldn’t all the climate conferences have an impact? When I went to university many years ago, I’m sure that professors weren’t haring off to conferences all over the world every few weeks. Now they see each other a lot and that must surely smooth things over.

“We maintain that the actions of researchers show that data mining is a necessary part of econometric inquiry. We analyse this phenomenon using the analogy of an industry producing a product (econometric analyses). There is a risk of selective reporting as Mayer indicates but we argue that other researchers (competition) will ensure that the sensitivity of truly important findings is checked. Hence, initial researchers have an incentive to analyse sensitivity from the beginning and so produce a quality product. Some suggestions are made towards encouraging this process. The ‘general to specific’ approach to data mining as promoted by Hoover and Perez can be valuable but it is premature to eliminate other strategies.”

The bolded part immediately elucidates the problem brought to proxy climatology by Michael Mann. His obscure language in publication, his secrecy concerning methodology, and his continuing obstructionism has frustrated the “competition” that should have had the opportunity to check his results. (One might say the same about Phil Jones.)

This analogy has the further benefit of putting Wegmann’s social network analysis in its proper light. That is, by roping most proxy workers into a collaborative relationship, Mann (deliberately or not) ensured there is no competition at all. He produced in the proxy community the business-equivalent of an interlocking directorate. The result was monopoly and restraint of trade. The “incentive to analyse sensitivity from the beginning and so produce a quality product” was entirely short-circuited.

As a result, the product was not quality, as you and Ross have shown.

Unfortunately for us all, Mann’s product had an addictive ingredient that has enslaved the minds of his consumer base. They are very angry about being exposed and very threatened by the incipient removal of a product that eases so much (psychological) pain, and are going to great lengths to keep it in the market.

A. R. Pagan and M. R. Veall (2000) “Data mining and the econometrics industry: comments on the papers of Mayer and of Hoover and Perez” Journal of Economic Methodology 7(2) 211-216