News for hackers. Stuff that matters.
Re:Chase had ads on 400K sites then on just 5K and got the s... by gwern on Saturday February 02, 2019 @11:44PM (#19067910)
I review some of the statistical/experimental papers in https://www.gwern.net/Ads . For example, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2367103 observes that given plausible effect sizes / variances, you can't run an advertising experiment which will pay for itself; https://www.kellogg.northwestern.edu/faculty/gordon_b/files/... demonstrates that even using absurd amounts of detailed tracking data from FB, you still can't recover actual causal effects.

There are no comments on this comment yet.