2024 Presidential Election - Take II

[Edgy MD]

Polls certainly are statistics, as are many models and estimations.

[rchurch314]

they're not like coin flips though. They're opinions. These polls are not revealing facts the way you'd calculate odds in a game of chance for instance. Nor is it sequential. "trump lead in July but now Harris leads in October" is an incorrect way to view information that has not changed, because it's information that has not happened yet. Hence, opinion. It's good for writing stories, but it really tells us next to nothing about what's going to happen because the future hasn't happened yet.

Oh, there's also this.

SOURCE Iowa Poll of 808 likely voters, Oct. 28-31. Margin of error: plus or minus 3.4 percentage points.

Like, the lead isn't even within the margin of error, and we're talking about like 24 people difference. Nearly 2 million people will vote in Iowa, but we're going to predict it based on a dozen or two?

Note they rounded the total numbers but went precise on the MoE. So that 3% difference could easily be 2.6 with rounding, or like 21 people, meanwhile the error number is 27 people.

[Edgy MD]

I certainly wouldn't suggest that polls are coin flips.

But polls are a recording of information that has "happened."

[whippoorwill]

I just texted this to a spam political texter


I decided when I got my 50th unsolicited text about voting that I was changing my vote.
Congratulations

Lying of course. But maybe in 2028 they’ll find something else to do on a nice autumn afternoon

[Edgy MD]

Here's a reason for MAGA-resistors to be optimistic about Pennsylvania — the Keystone State electorate are the ones that sent Dr. Oz packing in favor of a guy recovering from a brain injury, and who seems to hate dressing for work.

[whippoorwill]

Lol. Yeah he’s a nice guy though even though he looks like a thug.

My daughter in law and some of my friends have spent a good deal of time with him and he seems sincere

[Chad ochoseis]

they're not like coin flips though. They're opinions. These polls are not revealing facts the way you'd calculate odds in a game of chance for instance. Nor is it sequential. "trump lead in July but now Harris leads in October" is an incorrect way to view information that has not changed, because it's information that has not happened yet. Hence, opinion. It's good for writing stories, but it really tells us next to nothing about what's going to happen because the future hasn't happened yet.

Oh, there's also this.

SOURCE Iowa Poll of 808 likely voters, Oct. 28-31. Margin of error: plus or minus 3.4 percentage points.

Like, the lead isn't even within the margin of error, and we're talking about like 24 people difference. Nearly 2 million people will vote in Iowa, but we're going to predict it based on a dozen or two?

Note they rounded the total numbers but went precise on the MoE. So that 3% difference could easily be 2.6 with rounding, or like 21 people, meanwhile the error number is 27 people.

They're not just someone's opinion, and they're not pure science either.

The pollsters spoke to some number of people. They got 808 answers. Of those 808 answers, 47% of them said they planned to vote for Harris and 44% said they planned to vote for Trump. These are facts.

"Margin of error" is possibly the most misunderstood phrase in political discourse. The margin of error, which is really just a statistical convention, is two standard deviations. If your sample is unbiased, there is a 95% chance that the right answer falls within two standard deviations of the answer you took from the sample.

The statistical theory states that in a situation where you're taking a sample of a massive group of people and they're choosing between two options where about half will choose one and half will choose the other, two standard deviations in percentage points is 1 divided by the square root of the sample size. 1/SQRT(808) is 0.035. This is close to the 3.4 points they're using.

What the 3.4% margin of error means is that, assuming there is no sampling bias, there is a 95% probability that between 43.6% and 50.4% of Iowa voters support Harris.

That's the math. That shows that the 21 person difference isn't that much of a problem statistically. Or, at least, the size of the problem can be measured mathematically, and it's 3.4 points one way or the other.

You're correct that they would have done better to get a decimal point extra of precision in the percentages.

That's the fact part of polling. The opinion part is how to be sure the poll is unbiased. If you telephone people at random, but it's 1952 and there are still a large number of people who don't have telephones and that demographic breaks heavily for Truman, your poll is going to be inaccurate no matter what the math says. If it's 2008 and you're still conducting polls by calling land line phones, you're going to miss the younger voters who have gone 100% cellular and went heavily for Obama.

Or, if you believe Nate Silver, you might just pitch a poll that doesn't produce results that don't conform to what the other polls are saying.

The Des Moines Register is respected because they generally do a good job of avoiding biases like this.

My long winded point here is that the poll isn't perfect and it isn't conclusive. But it's based on data, not bullshit.

[batmagadanleadoff]

[whippoorwill]

Explain please

[batmagadanleadoff]

Explain please

It's a series of cards I first posted almost exactly four years ago, just before the 2020 election, when the Hitler was the actual sitting U.S. President. They're a continuation of the cards posted at page 17 of this thread.

[MFS62]

Can I claim I'm schizophrenic and vote twice?
If I claim to be bi-polar, the votes would cancel each other out.
Later