Statistical significance.....

Gelman and Stern published a paper that every social scientist employing statistical analysis should read. This is, of course, not the only paper that every social scientist employing statistical analysis should read, but it makes an important point that is almost certain to be lost on those who merely read the products of social scientists that use statistics. Ok, ok, so statistical significance is not the holy grail. That is not the point. The point is that differences in statistical significance often cannot be statistically distinguished from one another in statistical terms. And this should be troubling. Troubling above and beyond the other oft cited issues with statistical significance.

HIV, Circumcision, and Ethical Experimentation

In a fascinating story, the BBC reports that a randomized trial of circumcision halves the rate of HIV infection among heterosexual men in Kenya and Uganda consistent with previous evidence from South Africa. The story laudably notes that circumcision should not be thought to justify sex without protection and this point should not be overlooked. What I find fascinating is the consequence of a pilot study for experimentation. In short, the US NIH had to stop the trials early because it deemed a continuation of withholding circumcision from the control group unethical. While in this case, I cannot disagree; it would be morally reprehensible to continue. That said, I fear that this creates a dangerous experimental precedent for treatments with consequences that are less well known. Were it not that circumcision is a practice that dates back centuries, but instead was a pharmaceutical, say Vioxx or some relation, might we be inclined to engage similar moral concerns and ultimately support the wide employment of a treatment that leads to forms of longer term harm that might have been known were the study to have continued. In short, in this case it makes sense. What I hope is that researchers remain skeptical because a part of the justification for rigorous trials is the hope that unintended consequences may be discovered.