Funnel Plots and Publication Bias

I participate in the same behavioral genetics seminar attended by Tom Bouchard and his colleagues

http://www.psych.umn.edu/courses/fall10/psy8935/default.htm

from the University of Minnesota Twin Study most weeks during the school year, in an effort to learn more about the latest research on human behavioral genetics. The larger lesson here is that most initial studies of genetic associations with ANYTHING of interest in human beings get a lot of press from their gee-whiz value. Human beings have a cognitive bias toward assuming that their differences with other human beings are a matter of nature rather than nurture, if the other person can be deemed to be inferior. (The fundamental attribution error in psychology

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

results in most human beings most of the time concluding that their own faults and foibles are an outcome of the press of circumstances, while the other person's faults are mostly an outcome of the other person being a bad person by nature.) In fact, what human genetic research has shown over and over and over again is that any one genetic influence (from any one gene) on human behavior is very small indeed, usually all but undetectable even in large sample sizes, and that many human characteristics thought to be highly controlled by genes (highly "heritiable") are exquisitely sensitive to environmental influences.

http://people.virginia.edu/~ent3c/papers2/Articles%20for%20O...

http://people.virginia.edu/~ent3c/papers2/Articles%20for%20O...

http://people.virginia.edu/~ent3c/papers2/Articles%20for%20O...

It will take a long time for the general public and people who mostly read blogs to get their information about science to catch up with what the professional research literature says on this point, especially because human cognitive biases run against understanding the facts on this issue. But if the research program continues thoughtfully, we may eventually have opportunity to form a more correct understanding of just how much and also how little influence genes have on human behavior.

This is why clinical trial registries are so important. Unfortunately the last I heard they were a confusing mess, with many not actually providing the openness that was necessary for them to be effective.

What about "publish or perish"? If you spend a year doing a study, then I'd say you're going to publish the results, particularly if a lack of publications is going to mean that you're denied tenure or promotion.

That doesn't square well with the data here.

One possible explanation (and I'm open to others) is that the data shown here are not necessarily representative, but were selected to illustrate a particular idea.

This particular plot appears to show evidence of winner's curse rather than publication bias.