This seems to be related to the winner's curse in Auction theory. There's gotta be something related in the literature on the estimation of small treatment effects.
A question for stats/metrics folks. Is there a name for the following concept in hypothesis testing? Thanks!
X is the largest value of the treatment effect (dif in means) at which, to reject the null of no effect, you *must* draw a sample with mean>X? #Econometrics#Statistics
Intuitively, even if you have power to reject the null, under small enough effects you will always overestimate the true effect when you reject. Even if you make the right decision, your learning is wrong.
I am really upset that @GovKathyHochul pulled out of congestion pricing in Manhattan. Bad policy, and also a sad last-minute change of course. That's the tweet.
I am boarding a plane, and one of the attendants complains: " I swear to God, every time we make the overhead compartments bigger, people bring bigger bags!" People respond to incentives after all 😅
@nachocigliutti@ashdgandhi@paulgp My papers are currently only available as my dissertation, but me and my co-author studied pricing and reviews on Steam. Here is the link:
https://t.co/JdK0InSNXp
@tyleransom@JuliaLanguage Julia is great. I like the combination of "clear" and "powerful." I mostly use R day-to-day, but when I need to estimate something harder than a linear regression, Julia is my go-to!
GitHub's #Copilot knows you should use the full sandwich expression for the variance-covariance matrix of a maximum likelihood estimator in case the model is misspecified. Thanks to all the folks on #EconTwitter and beyond who did the right thing in their code over the years!