Hacker News Reader

Comment by westoque | Hacker News Reader

> ² According to the raw unweighted data, we surveyed 1,296 teens in the US and 1,309 teens in the UK and asked them if they experienced a range of feelings or experiences in the past month ... then asked if the feeling started on Instagram

I'm not a statistician but isn't 1,296 (US teens) + 1,309 (UK teens) such a small sample size to make these conclusions with?

A quick Google shows:

> In 2019, approximately 21.05 million young people between the ages from 15 to 19 lived in the United States.

Source: https://www.statista.com/statistics/221852/number-of-youth-a....

EDIT: Added link to US teen numbers.

EDIT 2: Thanks for the replies and further reading it's more about confidence and margin of error.

View on HN

No, it's plenty to measure the effect sizes reported. The larger the effect, the smaller the sample necessary to detect the effect. In this case the sample is large relative to the effect.

A bigger concern would be that the results don't generalize outside the US and UK, and my guess is that they probably don't.

I'd love to hear from the downvoters of this comment, of which it appears there are several. The comment reflects my understanding too, but I'm not an expert on study design/statistics and would like to know if there's something I'm missing.

I worked at Facebook around the time this data was collected, and read these results internally as part of my job. I analyzed similar surveys to this one and worked on doing things like designing experiments based on predicted effect size.

Maybe it's because I didn't say this originally. The sample size agrees with my intuition about how large a study should be for the sorts of questions it wants to answer. If the proportion of people giving an important answer were very small, say 0.01%, then the sample size would need to be much larger.

Yes, that roughly matches my understanding as well, which is why I was curious what the actual criticism of your comment would be, in case you and I are both missing something.

From the fact that even that neutral curiosity was downvoted, I'm inclined to believe that there is no such substantive criticism beyond "Facebook bad, me angry". Sometimes I forget how poor the quality of the average HNer has gotten.

You can get a UK election pretty much bang on with a similar sample size. It's not really about the sample size but how you handle the data.

Sample sizes are appropriate.

https://online.stat.psu.edu/stat200/lesson/8/8.1/8.1.1/8.1.1...

Reply on news.ycombinator.com