X pointed me to this atrocity:
The data on obesity are pretty unequivocal: we’re fat, and we’re getting fatter. Explanations for this trend, however, vary widely, with the blame alternately pinned on individual behaviour, genetics and the environment. In other words, it’s a race between “we eat too much”, “we’re born that way” and “it’s society’s fault”.
Now, research by Lazaros Gallos has come down strongly in favour of the third option. Gallos and his colleagues at City College of New York treated the obesity rates in some 3000 US counties as “particles” in a physical system, and calculated the correlation between pairs of “particles” as a function of the distance between them. . . . the data indicated that the size of the “obesity cities” – geographic regions with correlated obesity rates – was huge, up to 1000 km. . . .
Just to be clear: I have no problem with people calculating spatial autocorrelations (or even with them using quaint terminology such as referring to counties as “particles in a physical system”). I do have problems with this sort of gee-whiz reporting and the leap from an autocorrelation function to “it’s society’s fault.”
How about peer effects. I find those just as weak. I would love your take on the fight about the findings by Christakis and Fowler. I remember seeing the lengthy argument over those findings. Peer effects appear to be doing the same thing, except at a small scale (i.e. a friend’s fault I am fat).
Jonathan:
I’ve written a lot on this. Just search the blog for Christakis and Fowler.
Thanks as always!
at least one critic of the Christakis and Fowler work on peer effects has had most of his criticisms strongly undermined or refuted by other statisticians, and by experiments being conducted by this and other research groups. see: http://people.hsph.harvard.edu/~tvanderw/Papers/ValidSNTests_SPP.pdf
Anon:
I took a quick look at that paper and they’re all about testing the null hypothesis. I’m not so interested in testing the null hypothesis. Also I’m curious what they think about the analysis showing contagion of height.
This paper by vanderweele was a response to Lyons.
Christakis and Fowler already have responded to the paper by other critics claiming to document the contagion of height. See http://arxiv.org/abs/1109.5235 for more details, but, here is an excerpt: “At first pass, it would seem that height should not spread. Yet, in adolescents, it is not inconceivable that it might, and environmental factors explain a significant portion of the variance in height (around 20%) prior to adulthood [Visscher et al. (2006)]. To the extent that adolescent growth is, as is well known in the medical literature, influenced by exercise, nutrition, and smoking, it is entirely possible that an adolescent’s height could depend (to some degree) on the height of his friends, to the extent that they share smoking or exercise habits, for example. Moreover, adolescents with tall friends could (and perhaps would) again report that they are taller than they really are, or that they were gaining height faster than they really were – since, unlike the FHS-Net where height was measured by nurses, in AddHealth, it was self-reported. “
Anon:
Yes, I’m aware of that response. My point is that I’m not interested in testing the null hypothesis. The null hypothesis is just about always false, if only from measurement error, as Christakis and Fowler argue in the height example.
Lord. I’m sympathetic to the idea that the environment affects our behavior, but has it never occurred to these researchers that people are not trees and can in fact move, and that that should inform our inferences?
P.S. I should be more generous — it could be the person who summarized the research, and not the researchers themselves, glossing over this problem.
I wish we statisticians could avoid this sort of “gee whiz” reporting too (thinking of more than one piece in Significance, online and in print…)
What’s the distinction between “we eat too much” (#1) and “its society’s fault” (#2)? Can’t #1 be because of #2; how are experiments typically distinguish to between #1 and #2?
I have an issue with physicists pretending to be social scientists, but without bothering to read or learn from any of the social scientific research that’s already out there. To be fair, other disciplines do this as well, and some with less of an excuse. (I remember being in an interdisciplinary conference and having a well-regarded economist proclaim, “why doesn’t anyone study the distribution of housework within families?” As if there haven’t been upwards of 10,000 articles published on the topic in sociology and demography journals since the 1970s, if not before.)
It’s is amazingly tricky to move between seeing cool correlations and making inferences in non-experimental human subjects data. Usually, good work requires that you need to be very, very tricky to make it all come together as you have to account for the participants in the study making their own choices. Under those conditions, even big effects (like smoking and lung cancer) can take a while to be established to everyone’s satisfaction.
The other worry I have is people not checking their modeling system. If the system of people was like a group of gas particles, than the obesity clusters might be very interesting. But assuming people move at random (or that they are exchangeable and are not different for a myriad of historical, cultural and genetic reasons) is odd.
But the correlation stuff, itself, is kind of cool. I just wish they’d tempered the conclusions.
It’s like the author of the physics blog post didn’t even read the article. The authors of the article never use the term particles to describe their methods. They look at economic indicators, but never mention society as a cause of obeisity in the article.
They could have chosen better maps though, terrible projection and the 3D effect isn’t necessary.
I would suggest to read the original source article before arriving to your own conclusions:
http://arxiv.org/abs/1202.6345
I took a look and I don’t buy it. The graphs are cool but between the aggregation and the discretization, I can imagine all sorts of artifacts can arise.
Andrew:
I do not think that the point of this discussion is to learn whether you “buy” it or not. Firstly, as a manner of good modals, you should never call people: “Those darn physicists”. Secondly, if you pretend to be a scientist, you should read and fully understand the original scientific article before expressing your opinion publicly. Specially in this case, since the original study does not claim causality from correlation functions. Please read the abstract and the full article before expressing your opinion, which appears to be highly biased. After reading the article, I would then suggest to remove this “atrocity” and unscientific discussion from this blog.
George:
Of course the discussion is about whether I “buy” it. But, just to be clear, this is what I wrote in the blog entry above:
The “atrocity” I’m referring to is the reporting.
Finally, there is no requirement that this blog be limited to scientific discussions. We also discuss literature, sports, even zombies!