Hacker News Reader

Comment by graphe | Hacker News Reader

>The study was published this week in the October issue of Autism Research and created with collaborators from the University of Toronto. The research participants -- employed individuals, 33 with autism and 34 neurotypical -- were asked to weigh in on hypothetical scenarios involving everything from inefficiencies to inequalities to quality concerns.

It's a survey of 67 people. Could have been online even. Nothing I'd take seriously.

View on HN

Sample size is a red herring. If the difference is big and stark enough, you can detect it even in a MUCH smaller sample than this. You only need a big sample if the difference you are looking for is subtle.

Mind you, this is assuming that the sample has been taken in an unbiased manner (nigh impossible in a voluntary survey like this) and that any conclusions you draw are applied only to the population you took the sample from (e.g., if your sample is undergrad psychology students at a particular university, your results apply only to undergrad psychology students at that particular university, not to the general population).

The issue is not the sample size. It's the sampling method. Even if you do a study on 1 billion people, your are liable to get an incorrect result if you don't take the sample correctly.

Of course, the other issue here is that we are claiming something that the study didn't actually cover. The researchers did not look at whether the subjects succumbed to the bystander effect or not: they just looked at how the subjects answered questions. These are two extremely different things! This is bad experiment design.

It was the fact that it was a survey that made me throw the article in my mental trash.

Why is a study of 67 people not serious? If their sampling is solid it should be more than enough.

It's a survey. Ever took one?

ASD is very dynamic. It is not linear spectrum, think of it more like a polar graph. I would think 67 may be enough people to get an idea for one "hallmark" but certainly not all of them as a comprehension.

Many ASD-related surveys are not confirmed-diagnosed either, but only self-reporting as ASD.

I mean I get it, it'd be a colossal pain to verify all that medical data, but it also makes for bad science.

If you self report as autistic, you're probably autistic. Especially because, in practice, an autistic person is defined as someone who self-reports autistic traits to a health care provider.

Self reporting autism can be surprisingly accurate.

Reply on news.ycombinator.com