I have a sad reaction to this, which is realising that I simply don’t trust anything that these kind of researchers come up with, because I assume them to be left wing bigots.
Yeah, calling this an "effect size" is just nonsense, and it is alarming that educational software can get away with such poor statistical practice. I'm hoping this was just a student project.
I reckon East Anglia has the best beer in England. There are so many great local brewers. Adnams is down the road in Southwold and keep putting out brilliant, innovative beers. Then there's Lacons in Great Yarmouth and a host of small names, from Nene Valley Brewery to Mr Winters. If you're in Norwich, check out the Trafford Arms: it's in a nondescript-looking building which was rebuilt after a WWII bomb, but it has a constantly rotating playlist of brilliant ales and a landlord couple who really know their stuff.
There are many ways that free and competitive markets can fail other than behavioural economics! Monopoly, informational asymmetry, externalities… All of these are plausibly pervasive.
Alternative take that was expressed in the other post on this: census data users found DP extremely hard to work with, and viewed it as an imposed solution from the ivory tower. I wonder if any user could chime in on this.
The authors do address this issue, by reweighting their treatment and control counties on observable covariates. But I agree with you that this isn’t the causally watertight research design that economists usually strive for.
It might be worthwhile using local lightning strikes as an instrument for 3G coverage. Others have done this, but not for fertility afaik. But the lightning strike data costs about $1000.
I’ve been in cities with inadequate street lighting, and driving in them at night is terrifying. Car lights are not an adequate substitute on a busy road. I agree that in small towns and the country, street lighting is unnecessary.
Parent’s point was that many many people will get much more than $200 value from the “expensive” model. Sure, a Bihar farmer won’t, but even an Indian software developer may easily do if he or she has Western clients.
Right, but what I meant was: the other tests that the article says are used for definitively proving discrimination are equally bad, and subject to the same objection. Just substituting “one standard deviation“ or “statistical significance“for “80%“ doesn’t fix the fundamental problem here, which is that there are unmeasured confounders.
> Since the 80% test does not involve probability distributions to determine whether the disparity is a “beyond chance” occurrence, it is usually not regarded as a definitive test for adverse impact. Instead, other statistically significance tests, such as the standard deviation analysis, may be used for this purpose.
But then my question recurs: isn’t this a ridiculous way to measure discrimination? It’s assuming that the only thing that differs between the different ethnic applicant pools is their ethnicity, which is essentially never going to be true.
> To measure adverse impact, we apply the EEOC’s “four-fifths rule,” which flags a position when one group is recommended at less than 80% of the rate of the most-recommended group
That seems like a nonsensical way to measure racial discrimination. What could justify it?
我的子包：https://wyclif.substack.com。
我写过一本书：https://www.wyclifsdust.com。
davidhughjones at gmail