4 ms·

I'd love to see a massive project to test every child in the nation. It's hard to plan without good data. Once we know who is the worst off, we can better prote

• greedo 5y ago
  I'll believe that when I see it. We can't even test for a virus that's killing 400K/year. Nor can we provide safe drinking water for a city that's been stuck with crappy water for years and well publicized to boot.
• dredmorbius 5y ago
  Universal testing is very expensive and provides minimal additional data. That's not an argument for no surveillance, it's an argument for reasonable surveillance. Modern statistics is based around the principle of random sampling for very good reasons: it works. With a small but properly-conducted and truly random sample, you'll gain a tremendous level of insight over a large domain. The additional resolution increases with the square of the sample size. If you want twice the resolution you need four times the sample. Expanding a sample 10x only buys about 3.3x reduction in error. But your costs are up 10x. In the case of environmental toxic exposure, you'd want to conduct stratified sampling that focuses on known problematic areas (high population, high exposure), but also includes areas in which exposure risk is thought to be low. For specific treatment, you'd likely need to test individuals. That's not the same as determining where the problem is. And ultimately remediation and prevention is hugely more effective that retrospective treatment. In the case of lead, once there's been exposure, harm has already been done.