That’s captured in the margin of error. Setting aside the validity of their argument, a few hundred people in an online poll is all you need to make population-wide statements.
Margin of error assumes the sampling is random and representative, which of course is very hard to achieve in practice and especially with a laughably small sample size. Not really some magic number of truth as you’d like to present it.
You can start making statistical conclusions from as few as 3 samples and from around 30 you’re only getting square root improvements in margin of error. 600+ respondents is more than enough if there’s no selection bias, which I’d trust YouGov to be able to handle.
That’s captured in the margin of error. Setting aside the validity of their argument, a few hundred people in an online poll is all you need to make population-wide statements.
Margin of error assumes the sampling is random and representative, which of course is very hard to achieve in practice and especially with a laughably small sample size. Not really some magic number of truth as you’d like to present it.
You’re right about keeping an eye on sampling but you don’t know much about stats if you think 600+ respondents isn’t enough to draw any conclusions.
You can start making statistical conclusions from as few as 3 samples and from around 30 you’re only getting square root improvements in margin of error. 600+ respondents is more than enough if there’s no selection bias, which I’d trust YouGov to be able to handle.