Lies, damn lies and statistics. The Emory study cited by that first article (http://cjlf.org/deathpenalty/DezRubShepDeterFinal.pdf http://cjlf.org/deathpenalty

Lies, damn lies and statistics. The Emory study cited by that first article (http://cjlf.org/deathpenalty/DezRubShepDeterFinal.pdf http://cjlf.org/deathpenalty/DezRubShepDeterFinal.pdf) doesn't inspire me much. Show me an observational study which the researchers claim "eliminates the bias arising from unobserved heterogeneity", and I'll show you a study that was conducted by overconfident researchers who're overselling the strength of their evidence. And no, after reading the paper I'm not personally convinced that they did deal with unobserved heterogeneity. I'm not even convinced that the authors fully understand what the unobserved heterogeneity problem's really about. They seem to think that it's something you can whisk away by just throwing a bunch of covariates at your data, which is not how it works. If they mentioned their conditional independence assumption, let alone attempted to defend it, I missed it.