Hacker News Reader

Comment by 0xcafecafe | Hacker News Reader

Interesting. Can it also have something to do with an extremely small sample size? These are both rare conditions so intersection might be even rarer.

View on HN

Commenter above addresses this. Considering the prevalence of each condition, if they were independent then we would expect that 2 out of 10,000 people should have both together. The fact that zero cases have been found suggests that there is either a diagnosis issue, or that the two conditions are not independent.

That was an arithmetic error. It’s 2 out of 1,000,000.

The total 620 is actually correct again: 0.0072 times 0.0003 times 311591917 is 673.04. I suppose the 53 person discrepancy is due to rounding towards 0.03% but using a more accurate number for the actual computation (e.g. 0.0277%).

You can’t just throw extra digits around like that. It’s a bit unusual and jarring to see the population of the US quoted as “311591917” even if that is the census figure, because such a high-precision number would only be true for about a minute and you don’t know which minute.

Chill please. I used the exact number from the paper as quoted somewhere else in this discussion. Which seems adequate when saying that the authors final number (sixhundredsomething) is most likely correct even if the intermediate value "0.02%" is obviously wrong (and probably just a mistake in the print).

If you have questions as to which point in time this number refers to, I kindly refer you to the original paper and/or the authors.

No need to be rude. Sig figs are important, I reworded the comment a couple different ways to make the comment polite, and I’d appreciate the same courtesy. The use of an exact number for the US population is not justifiable, whether you can cite the source for it is not germane.

620 blind schizophrenics should be enough for anyone.

That sounds far more reasonable. Still, that leaves hundreds of cases that either are not being diagnosed or simply fail to exist.

Reply on news.ycombinator.com