[Perdogg]

638 gives you a MoE of 2.31%.

MoE = 1.26 X sqrt((R% X O%)/(n-1))

= 2.31%

[Hawthorn]

>> MoE = 1.26 X sqrt((R% X O%)/(n-1)) <<

Are you sure? I thought the MOE was just two sigma, that is, the 95% confidence intervals.

[Principled]

1/sqr rt 638 is 3.96%

that’s how i’ve seen it on other threads

[Outraged At FLA]

I understand the formula, but you still need a larger sample, or you could just poll 10 people. The larger samples just seem to be more accurate over time.

[verum ago]

That formula isn’t valid for this poll because it has non-random sampling. They report a margin of error of “about 4%.”