Conversation

Twitter polls are no better than noise, they said. Concluding anything based off twitter polls is laughable and meaningless, they said. Real research is the kind that gets published in academic journals, they said. (look at that stable 2.2%!)

238

Replying to

The polls are showing an accurate census of people who see your tweets. The real question is whether that population is representative of any larger groups.

Aella

@Aella_Girl

Replying to

@Yozarian22

i feel like everybody just learned that 'is it representative' is a button they can push and so they just push it without understanding why and when and how representation affects interpretations

5:24 AM · Aug 15, 2022

Replying to

I think it depends a lot on the question. For mundane things like if you've ever caught covid, I'd guess your sample has a good chance of matching. For something like your poll on body/face preferences, I'd expect your fans to have skewed preferences.

Replying to

and

Isn't that a fair question? You could even prove it doesn't apply to your polls, like having some other big account re-run one poll. Or try yourself to re-run a poll that's been published trying to control for selection bias. At worst, you'd find out how to calibrate.