First the link to this week’s complete list as HTML and as PDF.
***
So half of all responding scientists can’t reproduce their own results and over two thirds those by others, but less than a third accept they might probably be wrong, according to Baker? What’s going on here? If I repeatedly let go of an apple in midair and it refuses to fall, shouldn’t that give me pause? Most results in the human sciences are of this all or nothing kind and not quantitative. It is not as if non-reproduction meant not getting the claimed constant but another slightly different one or not finding a strictly linear relationship but a slight curvature inside the error ranges. Such later corrections would indeed not invalidate the earlier result, but these are emphatically not, what we’re talking about here.
***
As a result Okbay et al. is utterly meaningless. With an effect of 0.02 standard deviations, an R2≈0.02, and a negligible total effect even when taking them all together, the significance of those variants is purely statistical. If that. The statistics the authors are using is obviously that for a single test, not a massive data dredge. Admittedly their cut off point is not 5 % but 50 ppb, but then the number of genes tested must be of the same order of magnitude as the number of subjects. (I’m not sure how many genes there are in humans, not even sure if there is an accepted definition on how to count them. When I went to school, a gene was a DNA triple encoding one amino acid, but that definition seems no longer to be valid.) And incidentally the number of full authors, not acknowledged contributors, is not far off.
They do make a good point abut the effects of their best variants being stronger than expected purely from statistical outliers alone. But what does that mean? Does it really imply causality? I think not. Tiny effects like those (and even far larger ones) are usually the result of confounders, statistical covariance going back to a common as yet unidentified origin. The effect has certainly been proven not to be genetic in origin, at least not in the simple accepted sense of the term.
And lastly there is the question of what years of schooling is a measure of. Sure, there is a difference between those who leave school at the earliest legal opportunity and those who carry on to finish a university degree, but inside each subgroup the rank is reversed. The best ones are those achieving their goal in the shortest possible time. This statistical fallacy is well known to generate spurious nonsense in many other areas too.
***
Blaser and Palmer confirm a point I have been making for ages: Poisoning and annihilating your symbiotic microbiome can’t be healthy or beneficial.
***
Nitrogen isotope studies are only possible in protein tissue, which often is not preserved, and carbon isotopes don’t tell us much. So another isotope preserved in apatite and capable of elucidating diet, as claimed by Jaouen et al. for zinc, seems to come right on cue. Unfortunately their claims are not upheld by their data. While there is a slight shift in averages, the spread of the data for herbi- and carnivores overlaps completely and there is no predictive value to be had at all. Kudos for the nice try, but that seems to be about it, shame really.
***
I too had succumbed to the fallacy, that if traditional genetics does not tell all the story, then Lysenko must have been partly right after all. No, you can contradict something wrong and still be totally wrong yourself, as the new biography by Graham and the review by Meloni make a point of pointing out.
N.B: It surely is the pose as much as the face itself, but his portrait on the book’s title page uncannily reminds me of Rudolf Steiner.