Zum Seitenende Übersicht Artikel Home & Impressum
First the link to this week’s complete list as HTML and as PDF.
Beyer et al. claim to have measured the frequency of a 10-MHz-wide transition line to a precision in the order of kHz. They also claim all other measurements but one not only to be wrong by six sigma, but clustering as tightly as they do also all to be suffering from the same, unexplained systematic error. It’s not of course impossible that they’re right, but I wouldn’t correct all the textbooks just yet.
What does Schunck tell us? Not much. Once again we are not shown any data whatsoever. The errors around regression lines only tell us what other lines would fit the given data nearly as well, but say nothing about whether they are a meaningfully good fit in the first place. So what do we see? After some grouping, i.e. eliminating individual extremes, women, all women, are more visually attractive than men. More than that, in figure 2 the least attractive women are clearly still better looking than the most attractive men. Is there really nothing more than that? No, all regressions are calculated the wrong way round to even test the hypothesis. What was claimed was not that women become more attractive after marriage to a rich man but that beautiful women make the wealthiest catches. So physical attractiveness would have to be treated as the independent variable to yield any statement whatsoever – and the noisier the data are (we’re not shown any, but can reliably assume a lot of noise) the more important that choice becomes.
“And this is science?”[R. P. Feynman]
Academics of my generation tend to make fun of required courses in scientific writing for graduate students. Students at that level used to be expected to have read enough articles to pick up how it’s done by themselves. Reading Tinnermann et al. I’m beginning to see the necessity. Even after closely reading both the article and the supplement I’m still uncertain about their exact procedure and what, when and in what order they measured. What exactly does a
“trial” in fig. 1d entail?
Some things do become clear though. Contrary to what is claimed this is no nocebo effect as commonly understood. In the training phase real differences between placebo and control were artificially introduced, so this is really a behavioural learning experiment. Secondly looking at the four outcomes, cheap and expensive placebo plus two controls, three display nearly identical progressions in supplementary figure S1d,e – the single outlier, the result hinges on, is the control to the expensive placebo.
Mitteroecker et al. describe a well known engineering problem. Quite often the environment near an optimum is not symmetrical but rather displays a steep decline towards one side. An example I’m quite familiar with is the mixture setting of a combustion engine, whose optimum lies very near to the lean limit. To provide for variation and shoddy maintenance you either opt for reliability and set all engines a little too rich or you accept a certain number of bad failures to make all the rest that much better. It seems that nature prefers the thinking of engineers and the greatest aggregate benefit while salespeople accept mediocrity to avoid the publicity of highly visible individual failures.
If technology were made not for morons but for competent users, i.e. those willing and capable of performing a Caesarian if necessary, a lot of things out there could be much better. Of course technology also offers the option of individual tuning, something that neither nature nor mass production can emulate.
Whatever your stance is on climate change and its causes, if mitigation comes as the cost-effective by-product of improvements that are worth having anyway, you won’t go wrong in implementing them. This is just what Griscom et al. seem to propose, although they are a bit low on detail beyond a few rather general statements.
I don’t normally think a lot of the yellow press and only use its referrals to primary sources, I would not otherwise have found. The case of Wagner on Seddig et al. is an exception though – his comment is much better than I could have done, so I’ll just let it stand.
Zum Anfang Übersicht Artikel Home & Impressum