{"items": [{"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/784943968872?comment_id=784950875032", "anchor": "fb-784950875032", "service": "fb", "text": "I think you said this in the post, but just to put it a different way, the important difference between the distributions is the fatter tail in the Pareto distribution. For the Gaussian one, probability drops off so quickly, that very soon better interventions are so rare that aren't worth looking for.
Much of the difference in shape (eg Gaussian being symmetric) is irrelevant to this point, so the example is unnecessarily confusing. You could use a Gaussian centered on 0 where everything below 0 is cut off, and the relevant difference would remain.
(I was surprisingly slow to get this last night. Allison helped me.)", "timestamp": "1462418616"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/784943968872?comment_id=784950875032&reply_comment_id=784989382862", "anchor": "fb-784950875032_784989382862", "service": "fb", "text": "→ \"You could use a Gaussian centered on 0 where everything below 0 is cut off, and the relevant difference would remain.\"
I'm confused; that is what I used? That's what a half normal distribution is.", "timestamp": "1462448664"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/784943968872?comment_id=784950875032&reply_comment_id=784990770082", "anchor": "fb-784950875032_784990770082", "service": "fb", "text": "→ Oops, yes. I read and looked at the charts too quick (and that was in a footnote). Nice post. I was saying the same thing.", "timestamp": "1462449760"}]}