Hacker News Reader

Comment by criddell | Hacker News Reader

I would think that conference attendees would be fairly representative of commercial game developers.

I don't know enough about statistics to say how big the error bars would be here, but I'm constantly astonished by how well statistical methods can work with even seemingly small samples.

View on HN

The issue isn't with sample size. It is with sample method.

The population of gamedevs who go to a conference is probably not a random selection of gamedevs in general.

Its a sample size of 4000 and electoral polls are smaller than that 1k or 3k

> The issue isn't with sample size.

The issue isn't with sample size.

The sample is almost certainly biased, we just don't necessarily know how much and in what direction.

sample size is irrelevant if the selection criteria are poor.

If the sample size is large enough the selection criteria doesn't matter.

I guess that's technically true, but only if your sample size is 100%. :-)

There were examples of very large polls that failed spectacularly because of the badly chosen samples

https://www.math.upenn.edu/~deturck/m170/wk4/lecture/case1.h...

Take a sample of 5000 random people (an enormous sample size) attending the Republican National Convention and ask them which party they will vote for.

I would expect smaller game companies to be less well represented.

But considering how many developers don’t live where the conference takes place and/or can’t get there I have to agree with the top here and say this survey is severely limited.

it would indicate the better paid ones who can afford to g to industry conferences

Well, you can do a lot with a small sample provided that it's a random sample.

It seems like you'd have to know how conference attendees differ from the population of game developers you're interested in.

Reply on news.ycombinator.com