false positives
Incorrect matches in record linkage—linking two records that are actually different people.
As defined in Ran Abramitzky and Leah Boustan on Immigration Then and Now
Where it comes up · 7
- Russ Roberts And Tyler1 claim
- Tyler Cowen on the COVID-19 Pandemic1 claim
- Franklin Zimring on When Police Kill1 claim
- Ran Abramitzky and Leah Boustan on Immigration Then and Now1 claim
- John List on Scale, Uber, and the Voltage Effect1 claim
- 214 - A Conversation With Siddhartha Mukherjee1 claim
- 204 | John Asher Johnson on Hunting for Exoplanets1 claim
Most important claims · 7 of 7Search all claims for “false positives” →
- Transit signals can be mimicked by other astrophysical phenomena such as grazing eclipsing binaries, but work by Johnson and graduate student Tim Morton showed that at Kepler's photometric precision the rate of such false positives would be only about a tenth of what was feared, because Kepler's light curves traced the transit shape so cleanly that each was almost a textbook example—the earlier fears stemmed from a noisier era of jagged signals.204 | John Asher Johnson on Hunting for Exoplanets
- Linking algorithms face a tradeoff: very conservative rules (link only a uniquely-matched name with no near-birthdate competitor) yield small samples with few false positives, while looser rules yield larger samples with more false positives; the book reports only results robust across all these linking algorithms.Ran Abramitzky and Leah Boustan on Immigration Then and Now
- The single biggest US error was that for about 40 days there was no FDA-approved working test for the virus: the CDC's initial test worked but a faulty probe caused false positives when batches were scaled out to public health labs, while academic labs that had developed working tests (like Alex Greninger's at University of Washington) faced slow FDA/CDC licensing—creating a critical blind period during which the infection spread undetected.214 - A Conversation With Siddhartha Mukherjee
- Measuring wrongful police killings only by counting the 'handful' of unarmed people killed is the wrong metric, because many victims are armed with knives, bats, or blunt objects—or merely appear to have guns, sometimes holding radios—none of which actually kill officers, yet they are killed anyway.Franklin Zimring on When Police Kill
- There are five reasons (vital signs) ideas fail to scale: (1) false positives—the original result was an unrepresentative bad sample; (2) failing to know your audience—overestimating how many people the program can help; (3) failing to know your situation—mistaking an efficacy test for a scalable one; (4) spillovers—market-wide equilibrium effects and behavioral spillovers like moral hazard; and (5) supply-side returns to scale—whether the idea has economies, constant, or decreasing returns to scale.John List on Scale, Uber, and the Voltage Effect
- Cowen argues that with cases rising exponentially, the single biggest thing that can be done—given tests aren't available—is for people to interact with others as little as possible, intentionally putting the economy 'in a coma' to keep the case load manageable relative to healthcare infrastructure and avoid a Lombardy-style collapse.Russ Roberts And Tyler
- People should focus less on the caseload and more on the death count, because there appear to be an enormous number of false positives and uncertain numbers of asymptomatic positives, making the case data less reliable than the (unfortunate) death data.Tyler Cowen on the COVID-19 Pandemic