Assorted links

Regarding #1: man, that does seem like a pretty terrible methodology. Obviously there’s so much data in a set of brain scans that there are bound to be some correlations. The point is to pull out which ones are not accidental byproducts of a small data set.

In many computational fields you do this by having a “development set” (say, a random 50% of your brain scans) where you can do “dirty” things like find very specific brain locations whose activity or lack thereof cleaves the data into the two sets you want. From this you develop hypotheses about which of those regions are actually pertinent to the phenomenon you are measuring. Then you see if those regions are pertinent for the remaining 50% of brain scans.

Taken to an inappropriate extreme this can be just as meaningless a method (if you try 1000 hypotheses, some of them will be true by accident), but if you try a small number of hypotheses based on prior knowledge of the brain, any resulting “positives” are much more likely to be genuine.

I can readily believe that these brain-scan studies suffer from serious methodological flaws. I spent many years working with scientists as a statistician and bioinformatics programmer, and the pressure to generate a “significant” result is intense. Even very accomplished scientists have a hard time accepting that their most cherished correlations are poo-poo.

The best way around this is to establish (preferably in advance) a specific, objective algorithm for interpreting the data. Then you run the algorithm on both the original dataset and a series permuted datasets where cases and controls have been randomized. When “significant” results come up from the permuted dataset, it’s pretty easy to convince the scientist that the non-permuted results are bogus. If you can perform a few thousand permutations, it’s also a good way to get a credible P-value when the analysis is too complicated to compute one analytically.

Most nefarious is a problem I call the “fishing expedition” – the scientist privately examines a large number of hypotheses before presenting only the most “promising” ones to the statistician. Because you don’t know how many hypotheses were rejected, and usually the scientist doesn’t either since they aren’t even aware they’re doing it, there is no way to perform an appropriate Bonferroni correction.

That is why independent replication of the results is so important.

It will be obvious to the reader that I have been seeking not simply to describe and explain that bundle of changes which can be reasonably lumped together as a trend to financialization, but also to air my prejudices against it. There are alternative views, however.

One is as follows. The real world economy is growing. Its need for credit and insurance will obviously grow too. Even if it only grew at the same rate, given that financial markets are increasingly global, it is not surprising that the finance industry should grow lopsidedly in some countries (Britain, for example, where it employed 11% of the working population in 1980 and 22% today) rather than in others. The apparent overblown growth of the finance industry in the Anglo-Saxon economies is simply a function of the division of labour. Maybe, but that is hard to reconcile with the fact that the volume of exchange transactions is a three-figure multiple of the volume needed to settle world trade.

A second argument is as follows. As material technology so improves that a declining fraction of the labour force is needed to produce an ever-expanding volume of material goods, more and more income is devoted to the purchase of, and more and more people are employed in the production of, services. Business services expand as more and more complex specialties are provided through the market rather than in-house, tourism expands, entertainment expands, and naturally financial services expand too; more people have larger volumes of financial assets which they are anxious to have taken care of and the growth of the speculative element in financial transactions simply parallels the growth of casinos in the entertainment industry.

There is a clear answer to that. Most of the speculators are not betting their own money – the principals do not even get a buzz from the gambling – and the speculators do nicely, thank you, whether the principals win or lose. And when ordinary people who do not have millions in hedge funds are forced by the pension or mortgage system to make their own, inevitably speculative decisions – usually on the basis of tenuous information – it is often their own basic security they have to put at risk.

So I will stick to my prejudices.

Restaurants in Wexford |Nite Clubs in Wexford|