A word about polls, heading into the primaries:

[CTyankee]

Always look at the size of the poll. If the polling sample is small (say a 500 person poll), the margin of error (M.O.E.) goes UP. The opposite is true when the poll is larger.

This is a math rule, not my opinion.

[kimbutgar]

1. I am highly suspious of that wsj poll

I wonder what the sampling of repukes vs independents and Democrats were.

If I were a waging person it was done to give the orange maggot a cookie to calm him down.

[rogerashton]

2. True, but

1) It is the absolute size of the sample that matters for that purpose, not the relative size. Five hundred out of 300 million is just as good as 500 out of 1000.
2) Qualification: this assumes the differences are independent and that very rare events are not important. A poll of 500 taken at random will not tell us much about the voting preferences of billionaires or climate scientists.
3) In political polls, the main problem is with the fudges: how are likely voters determined, etc? Allowing for that, the surveys are pretty far away from random, even if the number is large.

[CTyankee]

4. I am not seeing your reasoning on #1.

I agree with you on 2 and 3.

[rogerashton]

5. That's what the central limit theorem says.

And the central limit theorem was the "matter of math" you referred to.

https://towardsdatascience.com/understanding-the-central-limit-theorem-642473c63ad8

[CTyankee]

6. Thannks for the link. It clears up everything!

[rogerashton]

7. It's a very tricky point.

Often confused in the press.

[Odoreida]

3. Always be skeptical of polls ...

... especially when they tell you what you want to hear.

[dsc]

8. I agree with you for the most part

but 500 isn't that small a sample and the indirect variation is to the square root of the sample size not the sample size. Thus a sample size of 2000 has a MOE half as big as one of 500.