"average person eats 3 spiders a year" factoid actualy just statistical error. average person eats 0 spiders per year. Spiders Georg, who lives in cave & eats over 10,000 each day, is the AMIS. we've decided to show quantiles of the spider consumption distribution instead.
tumblr deep cut, real OG online folks will recognize
notice how unlike the original meme, we do NOT advocate droppping spiders georg. he is a real part of the DGP, presumably! don't drop him! but, his presence makes the sample mean unsuitable / uninteresting for our purposes, so we need to rethink the analysis in this case.
also, a halfway careful analyst will usually figure out spiders georg is there, and may trim or switch to quantiles or w/e without needing us to tell them. But in our paper we find that sometimes spiders georg looks like a normal guy. then you might need us to find him for you.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
So some very smart folks have asked about how we would apply the AMIP metric to studies of rare events. This kicked off a discussion of what robustness checks are really for, and I want to take that set of questions seriously in this thread.
I think robustness checks mainly (ought to) function to illuminate how variation in the data is being used for inference, and we should then be able to discuss whether we think this is a reasonable situation and adjust our confidence in the results.
The problem is not that there is SOME change to which our analyses are sensitive -- of course there is, they has to be. If your results aren't affected by ANY change you make to the analysis, something has gone horribly wrong with the procedure.
Guys this paper is super important. Arnold, Hull and Dobbie are among the most careful applied econometricians we have, and the explosion of algorithmic decision making means this method -- and their finding of pervasive discrimination -- could hardly be more timely.
hey since we were discussing the other day how even just "select high contrast areas to thumbnail" is a racist decision rule given the history of photography AND since you guys only understand one language, i made this for u
Hello! Tamara Broderick, Ryan Giordano and I have a new working paper out!! It's called "An Automatic Finite-Sample Robustness Metric: Can Dropping a Little Data Change Conclusions?" arxiv.org/abs/2011.14999
Here comes the paper thread!!! Aaaaaah!!!
We propose a way to measure the dependence of research findings on the particular realisation of the sample. We find that several results from big papers in empirical micro can be overturned by dropping less than 1% of the data -- or even 1-10 points, even when samples are large.
Also the London Review of Books is better than the New Yorker send tweet.
1. we have Lauren Oyler, who do u have? Jia?? Sad. 2. we have a Nabokov bingo square so incendiary I'd get immediately permabanned if I tweeted a screenshot. 3. we have new hits from Anne Carson
I do love the New Yorker and 2 NYer articles - one about the Bostswanan diamond mines and one about the deep sea submarine - were 2 of the best things I read all year. but LRB is better and in your heart you know it.