tag:blogger.com,1999:blog-4266300416434206693.post4017474181425425042..comments2023-08-11T12:04:42.077+02:00Comments on The Church of Rationality: Unpacking Karl Smith on Experiments and Regressions (An Introduction to Causality and How to Measure It), Part ILemmusLemmushttp://www.blogger.com/profile/00917054221547240969noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-4266300416434206693.post-55218580620683671032011-03-21T22:23:02.490+01:002011-03-21T22:23:02.490+01:00I have no intention of scoffing at sophisticated t...I have no intention of scoffing at sophisticated techniques for observational data, but I do have the intention of pointing out some weaknesses of simple multivariate regressions (in the next installment), which I wouldn't put in that league.<br /><br />Both controls and randomized experiments are done in an attempt to increase the probability that an observed statistical association is, in fact, causal, so talking about the merits of observational techniques vs. randomized experiments means talking about causality, no? After all, establishing causality is randomized experiments' big claim to fame.<br /><br />But perhaps we should continue the conversation when my job is done.LemmusLemmushttps://www.blogger.com/profile/00917054221547240969noreply@blogger.comtag:blogger.com,1999:blog-4266300416434206693.post-6712852795357669572011-03-21T14:21:42.368+01:002011-03-21T14:21:42.368+01:00I;m interested to hear the rest of your unpadking,...I;m interested to hear the rest of your unpadking, but its important to note that I was making no statement about causality.<br /><br />My only point is that there is nothing magical about random trails. You can't simply say "we randomly assigned some to control and some to treatment" and then think your job is done.<br /><br />Nor should you look askance at sophisticated observational techniques simply because they do not have controls.Karl Smithhttp://modeledbehavior.comnoreply@blogger.com