Hacker News Reader

Comment by unparagoned | Hacker News Reader

That’s wrong. Only if the studies are powerful enough and are looking at the right stuff can you make any reasonable conclusions.

If the studies are powerful enough then that absolutely isn’t evidence of absence at all.

View on HN

Evidence can be strong or weak. Every positive study result is evidence of presence, usually strong evidence. Every negative study result is evidence of absence, usually very weak evidence.

And yet, people can be extremely vigilant about a positive correlation when it doesn’t fit their personal agenda, going as far as questioning methodology in even the least pragmatic ways possible; or, become accepting of the absence of evidence as evidence of absence when weak/no correlation is convenient to their own biases.

I'd say it's very sensible to assume absence up front before there's convincing evidence of presence. Sure, personal biases influence our evidence thresholds a lot, and often people demand unfair rigor that cannot be practically met. But the opposite is also true - sometimes people will believe in presence with zero evidence if they like it enough. The latter is far, far more dangerous than the former.

It’s one thing to believe in presence with zero evidence, and another to _withhold_ judgment when there’s zero evidence. I think you may be conflating the two.

Then again, perhaps we are simply two different types of people. I don’t believe that a complex system can be expected to stay the same when you introduce a new factor into it, while you prefer to adjust much later when effects appear or when evidence achieves total infallibility up to the academic level. I am risk-averse, you like taking it; I’d rather prevent, you’d rather cure. Stating that the former is “far, far more dangerous” is just begging the question though.

That's not right, a negative study often still can give evidence that something is true, but it just didn't meet the p=.05 threshold.

p<=0.05 is so easy to achieve for actual positive results (just increase sample size) that I don't think it's productive to treat any result with p>0.05 as positive evidence for any purpose other than deciding to rerun the study with a bigger sample size.

That's not true at all, that's a weird generalization that's patently false.

Take things like even certain physics propositions (e.g. discovering new particles) -- can take billions of dollars building a new reactor and years of experiments to find a result.

Anything medical or psychological also can take this result. If a chemical reduces your iq by 5 points over 10 years, it might not be able to do a study that finds that result on an academic budget.

Reply on news.ycombinator.com