Comment by FlyingSnake | Hacker News Reader

Comment by FlyingSnake | Hacker News Reader

>They ended up with a final sample size of 813 people.

I want BlueSky to succeed but this sampling bias is simply too much to ignore.

This comment (by nunobrito) from few days ago on a similar topic is best analysis of this topic.

> These news are awfully similar to click-bait stating "the science is settled" by grouping a small set of the group and then pretending it represents the whole. The paper failed both to identify the overall number of scientists using X or the cases where multiple platforms are used (most common scenario). Therefore the paper only seems biased on its best scenario or downright propaganda at its worst. > NOSTR and Mastodon should never be left out of any serious research.

https://news.ycombinator.com/item?id=44982510

If the poll was done _properly_, that sample size is _fine_; there’ll be a decent margin of error, but not as much as you might expect. 1k people is a fairly standard size for polls, with even very high quality ones rarely doing over a few thousand.

The real consideration was whether the poll was done properly.

What is the sampling bias? You dont explain what it could be, and your quote doesn't give any clues about what the bias could be.

The article itself talks about this self-selection/sampling bias due to a minuscule sample size of 813 people. Reducing “science community” to such a small sample is not convincing.

Self-selections, sure, that's a risk with any and all surveys and questionnaires, which must be mentioned.

But 813? Why wouldn't that be enough? Basic stats puts that at a very healthy number for most questions, and the researchers don't raise any questions about bias about the number.

> The paper failed both to identify the overall number of scientists using X or the cases where multiple platforms are used (most common scenario)

That would not indicate bias. It's merely beyond the reach of what was in the study.

Reply on news.ycombinator.com