On the idea that the polls got the election wrong

[YoUDeeMan]

La times poll

That was a NATIONAL poll that had Trump winning by 3 points NATIONALLY.

It wasn't predicting the electoral college. It was predicting the national popular vote. Clinton won the popular vote by 2 points. So, LA Times poll was off by 5 points.

Yup..another poll that was...WRONG!

[YoUDeeMan]

Thought I'd put this here:

"Take their first claim. Only liberal academics would believe that “voter-impersonation fraud” is a myth.

A recent study in the peer-reviewed journal Electoral Studies estimated that illegal aliens cast about 1.4 million votes in the 2008 and 2010 elections, and that their votes “likely gave Senate Democrats the pivotal 60th vote needed to overcome filibusters in order to pass health care reform and other Obama administration priorities in the 111th Congress.”

Here are some cases of suspected impersonation fraud discovered during just the month-and-a-half prior to the November election:

• San Pedro, Calif.: 83 absentee ballots were sent to different registered voters who all supposedly lived in the same small, two-bedroom apartment. If not for an observant neighbor, this case would never have been discovered.

• Pennsylvania: Democrat organization FieldWorks LLC was raided by Pennsylvania State Police on allegations that it fraudulently filled out registration forms for thousands of voters.

• Indiana: State police “believe there could be hundreds of fraudulent voter registration records with different combinations of made up names and addresses with people’s real information.“

• Chicago: An investigation by CBS-2 found that “119 dead people have voted a total of 229 times in Chicago in the last decade.”

• Virginia: After examining just eight of the state’s 133 counties, 1,046 illegal aliens were discovered to already be registered voters.

• New York City: In an undercover video, Democrats themselves were caught complaining about the amount of voter fraud created by Mayor Bill de Blasio’s decision to give out ID cards without checking recipients’ identities."

http://www.foxnews.com/opinion/2017/01/ ... =obnetwork

[cx500d]

That was a NATIONAL poll that had Trump winning by 3 points NATIONALLY.

It wasn't predicting the electoral college. It was predicting the national popular vote. Clinton won the popular vote by 2 points. So, LA Times poll was off by 5 points.

Yup..another poll that was...WRONG!

They weighted and sampled it appropriately per electoral vote.


Sent from my iPhone using Tapatalk

[JohnStOnge]

That was a NATIONAL poll that had Trump winning by 3 points NATIONALLY.

It wasn't predicting the electoral college. It was predicting the national popular vote. Clinton won the popular vote by 2 points. So, LA Times poll was off by 5 points.

Yup..another poll that was...WRONG!

Actually it was but not by much if you properly interpret it...even though I think it like other polls understated the margin of error. But let's go with their margin of error. It was 2.5 percentage points. And the point estimate was Trump by 3 percentage points.

As I've written before the margin of error refers to an individual estimate for one of the candidates. The margin of error for the difference between the two is about twice the margin of error reported. So the margin of error for the difference was about 5 percentage points. That means that you'd look at it and say it's somewhere in the range Trump by 8 through Hillary by 2. Hillary won by 2.1 percentage points.

Of course I think they got whacked by what I talked about: That margin of error is random sampling error and there are other sources of error in these political surveys. I really do think the pollsters ought to do some work to try to take those other sources of error into account and report margins of error based on that.

Also I think the news media need to stop reporting point estimates. What they need to do is report the ranges. The way they do it now gives the public a false impression. Right now they'll do something like "Clinton 46, Trump 43." What they need to do is "Clinton 43 to 49, Trump 40 to 46." When you do it that way you can clearly see that you can't really distinctly separate the two. You can see that it's possible Trump is ahead. It'll never happen but the media really do need to stop reporting poll estimates as single distinct numbers for each candidate.

BTW it's not exactly like that because the margin of error actually changes with different point estimates. It's at it's maximum at 50% then declines as you move "outward" from that. But we won't get into that for now.

[cx500d]

Yup..another poll that was...WRONG!

Actually it was but not by much if you properly interpret it...even though I think it like other polls understated the margin of error. But let's go with their margin of error. It was 2.5 percentage points. And the point estimate was Trump by 3 percentage points.

As I've written before the margin of error refers to an individual estimate for one of the candidates. The margin of error for the difference between the two is about twice the margin of error reported. So the margin of error for the difference was about 5 percentage points. That means that you'd look at it and say it's somewhere in the range Trump by 8 through Hillary by 2. Hillary won by 2.1 percentage points.

Of course I think they got whacked by what I talked about: That margin of error is random sampling error and there are other sources of error in these political surveys. I really do think the pollsters ought to do some work to try to take those other sources of error into account and report margins of error based on that.

Also I think the news media need to stop reporting point estimates. What they need to do is report the ranges. The way they do it now gives the public a false impression. Right now they'll do something like "Clinton 46, Trump 43." What they need to do is "Clinton 43 to 49, Trump 40 to 46." When you do it that way you can clearly see that you can't really distinctly separate the two. You can see that it's possible Trump is ahead. It'll never happen but the media really do need to stop reporting poll estimates as single distinct numbers for each candidate.

BTW it's not exactly like that because the margin of error actually changes with different point estimates. It's at it's maximum at 50% then declines as you move "outward" from that. But we won't get into that for now.

Wah waaaaah wah waaaaaah wah waaaaaaaah.


Sent from my iPhone using Tapatalk

[YoUDeeMan]

Actually it was but not by much if you properly interpret it...even though I think it like other polls understated the margin of error. But let's go with their margin of error. It was 2.5 percentage points. And the point estimate was Trump by 3 percentage points.

As I've written before the margin of error refers to an individual estimate for one of the candidates. The margin of error for the difference between the two is about twice the margin of error reported. So the margin of error for the difference was about 5 percentage points. That means that you'd look at it and say it's somewhere in the range Trump by 8 through Hillary by 2. Hillary won by 2.1 percentage points.

Of course I think they got whacked by what I talked about: That margin of error is random sampling error and there are other sources of error in these political surveys. I really do think the pollsters ought to do some work to try to take those other sources of error into account and report margins of error based on that.

Also I think the news media need to stop reporting point estimates. What they need to do is report the ranges. The way they do it now gives the public a false impression. Right now they'll do something like "Clinton 46, Trump 43." What they need to do is "Clinton 43 to 49, Trump 40 to 46." When you do it that way you can clearly see that you can't really distinctly separate the two. You can see that it's possible Trump is ahead. It'll never happen but the media really do need to stop reporting poll estimates as single distinct numbers for each candidate.

BTW it's not exactly like that because the margin of error actually changes with different point estimates. It's at it's maximum at 50% then declines as you move "outward" from that. But we won't get into that for now.

Wrong is wrong.

When you say someone is ahead in a poll, and that person loses...you were wrong. You don't get partial credit for being close.

To put it into a thought you can relate to, you don't say, "she got a partial abortion," or, "she got partially pregnant."

Of course, you can say she got a partial birth abortion...but we all know what that means.

[cx500d]

Actually it was but not by much if you properly interpret it...even though I think it like other polls understated the margin of error. But let's go with their margin of error. It was 2.5 percentage points. And the point estimate was Trump by 3 percentage points.

As I've written before the margin of error refers to an individual estimate for one of the candidates. The margin of error for the difference between the two is about twice the margin of error reported. So the margin of error for the difference was about 5 percentage points. That means that you'd look at it and say it's somewhere in the range Trump by 8 through Hillary by 2. Hillary won by 2.1 percentage points.

Of course I think they got whacked by what I talked about: That margin of error is random sampling error and there are other sources of error in these political surveys. I really do think the pollsters ought to do some work to try to take those other sources of error into account and report margins of error based on that.

Also I think the news media need to stop reporting point estimates. What they need to do is report the ranges. The way they do it now gives the public a false impression. Right now they'll do something like "Clinton 46, Trump 43." What they need to do is "Clinton 43 to 49, Trump 40 to 46." When you do it that way you can clearly see that you can't really distinctly separate the two. You can see that it's possible Trump is ahead. It'll never happen but the media really do need to stop reporting poll estimates as single distinct numbers for each candidate.

BTW it's not exactly like that because the margin of error actually changes with different point estimates. It's at it's maximum at 50% then declines as you move "outward" from that. But we won't get into that for now.

Wrong is wrong.

When you say someone is ahead in a poll, and that person loses...you were wrong. You don't get partial credit for being close.

To put it into a thought you can relate to, you don't say, "she got a partial abortion," or, "she got partially pregnant."

Of course, you can say she got a partial birth abortion...but we all know what that means.

It means she's not in Texas.


Sent from my iPhone using Tapatalk

[JohnStOnge]

Wrong is wrong.

When you say someone is ahead in a poll, and that person loses...you were wrong. You don't get partial credit for being close.

To put it into a thought you can relate to, you don't say, "she got a partial abortion," or, "she got partially pregnant."

Of course, you can say she got a partial birth abortion...but we all know what that means.

I started the post to which you are responding off by saying "Actually it was" in answer to the assertion that the LA Times Poll was wrong. I did say "not by much."

But here's another thing: When you look at that "margin of error" thing you expect the actual result to be outside of that about 1 in 20 times. So you know going on that even if everything is done perfectly it's going to happen. To determine that a poll is "unreliable" you'd have to determine that the true result is outside of the "margin of error" more than 5% of the time to a "significant" degree.

It's a very "fuzzy" thing if you really understand what's going on. My impression is that, unfortunately, few if any of the people we see while watching the news really understand what's going on. The look at the point estimates and don't have an appreciation for the "fuzzy" nature of what they're looking at.

[cx500d]

Wrong is wrong.

When you say someone is ahead in a poll, and that person loses...you were wrong. You don't get partial credit for being close.

To put it into a thought you can relate to, you don't say, "she got a partial abortion," or, "she got partially pregnant."

Of course, you can say she got a partial birth abortion...but we all know what that means.

I started the post to which you are responding off by saying "Actually it was" in answer to the assertion that the LA Times Poll was wrong. I did say "not by much."

But here's another thing: When you look at that "margin of error" thing you expect the actual result to be outside of that about 1 in 20 times. So you know going on that even if everything is done perfectly it's going to happen. To determine that a poll is "unreliable" you'd have to determine that the true result is outside of the "margin of error" more than 5% of the time to a "significant" degree.

It's a very "fuzzy" thing if you really understand what's going on. My impression is that, unfortunately, few if any of the people we see while watching the news really understand what's going on. The look at the point estimates and don't have an appreciation for the "fuzzy" nature of what they're looking at.

Your logic is also very "fuzzy"


Sent from my iPad using Tapatalk

[JohnStOnge]

I resurrected this thread because when I was killing time this past weekend for some reason I decided to look at what would happen if I went to the RealClearPolitics site and just looked at the single latest poll for each State taken during the period most immediately preceding the election. Included in that idea is that if two polls ended their sampling on the same date I considered the one with the latest start date to be the latest. So for example if one poll was conducted during November 5 and 6 and another was conducted just during November 6 I identified the one conducted just during November 6 to represent the latest period prior to the election.

And I think the results are very interesting.

The total tally is that when you take the "latest poll" point estimates they are 44-1-1 with respect to individual State + DC jurisdictions. The tie is because the "latest poll" point estimate for North Carolina is a tie.

The only "loss" in this context is Wisconsin. Had you just gotten up on the night before the election and all you saw was the single "latest poll" result for each State the only State that would have surprised you is that one. When you looked at the latest poll for Michigan you'd have seen Trump by 2. When you looked at the one for Pennsylvania you'd have seen Trump by 1. You'd have seen Trump by 4 in Florida.

[JohnStOnge]

I started the post to which you are responding off by saying "Actually it was" in answer to the assertion that the LA Times Poll was wrong. I did say "not by much."

But here's another thing: When you look at that "margin of error" thing you expect the actual result to be outside of that about 1 in 20 times. So you know going on that even if everything is done perfectly it's going to happen. To determine that a poll is "unreliable" you'd have to determine that the true result is outside of the "margin of error" more than 5% of the time to a "significant" degree.

It's a very "fuzzy" thing if you really understand what's going on. My impression is that, unfortunately, few if any of the people we see while watching the news really understand what's going on. The look at the point estimates and don't have an appreciation for the "fuzzy" nature of what they're looking at.

Your logic is also very "fuzzy"


Sent from my iPad using Tapatalk

I just happened to see this when I posted my last post. It's not a matter of my logic being fuzzy. It's the way survey sampling works. Even if you have a perfect situation, you have to consider random sampling error. The standard "margin of error" is a 95% confidence interval. When you say 95% confidence interval you are saying that you expect the true value to be within that interval 95% of the time when you go through that process. That means that you expect the true value to outside of the interval 5% of the time. That's 1 in 20. When you go through the process repeatedly you accept the idea that over a large number of repetitions the truth is going to be outside of the margin of error somewhere around once in 20 tries.

And THAT'S if you had a perfect situation where the only source of error is random sampling error. Political pollsters don't have that.