3 ms·

500 sounds like a sufficient sample size to me. How many samples would you need before accepting the results of a study?

• nordsieck 8y ago
  It's not necessarily any one aspect of the study, but all of the elements as a whole. 1. This is an observational study, not a double blind experiment. We know that these are basically garbage. In many cases it's unethical to do experiments, so observation is the best we have; that doesn't mean it's any good. 2. IQ pre puberty is highly volatile. The younger you test, the less your results correlate with eventual adult IQ. It may be possible to add say 100x more participants to compensate for the 2nd point. You can't add more participants to fix the first point; blacklisting confounders just doesn't work except in exceptional cases (e.g. smoking and cancer). The real way to do this study is 1. Double blind study instead of observation. This means along with everything they're already doing, they need to give the mothers a 16 year supply of water filters (presumably over time). Half of them would be inert. 2. Run the study until the kids are at least 16. Yes, this is more expensive, but the knowledge per dollar gained would be infinitely higher.
  • dpark 8y ago
    > but the knowledge per dollar gained would be infinitely higher. Would it be? It sounds like millions, potentially tens of millions, would be required to run this hypothetical study. And the likely outcome is that the effect is minimal.
    • nordsieck 8y ago
      >> but the knowledge per dollar gained would be infinitely higher. > Would it be? I meant that literally. 1/4x is infinitely larger than 0/x > And the likely outcome is that the effect is minimal. Everyone knew this was true before the study started.
      • dpark 8y ago
        I misunderstood. I thought you were saying the knowledge would be worth infinitely more than the millions this would cost.