Moore or Jones?

[AUDub 11,184]

Some positive signs If you're hoping to see Moore lose. 

http://fivethirtyeight.com/features/doug-jones-is-just-a-normal-polling-error-away-from-a-win-in-alabama/

Doug Jones Is Just A Normal Polling Error Away From A Win In Alabama

By Harry Enten

Filed under Special Elections

•  
•  
•  

Things seem to be going Roy Moore’s way. President Trump endorsed him. The Republican National Committee is back to supporting him. And Moore, who has been accused of sexual contact with women when they were underaged, has led by an average of 3 percentage points in polls¹ taken within 21 days of the Dec. 12 special Senate election in Alabama. The betting markets give Moore about an 80 percent chance of victory — roughly the same chance they gave Hillary Clinton just before the 2016 presidential election.

Before Election Day last year, we advised caution, however — polls aren’t perfect at even the best of times, Trump had an advantage in the Electoral College, and there were a lot of undecided voters. So what’s our advice heading into the Alabama election? Well, it’s the same — be cautious — but for slightly different reasons.

A look at all U.S. Senate election polls since 1998² shows that their average error — how far off the polls were from the actual election result — is more than a percentage point higher than the average error in presidential polling. Also, Alabama polls have been volatile, this is an off-cycle special election with difficult-to-predict turnout, and there haven’t been many top-quality pollsters surveying the Alabama race. So even though Moore is a favorite, Democrat Doug Jones is just a normal polling error away from winning. (Or, by the same token, Moore could win comfortably.)

[homersapien 11,452]

I just know that if Moore wins, I am so glad that I can hide the fact I am from Alabama.  (I'd remove the Auburn magnet from my car if I thought most people associated Auburn with the state of Alabama.)

Alabama has once again become an embarrassment to the rest of the country. 

[AUDub 11,184]

Just now, homersapien said:

I just know that if Moore wins, I am so glad that I can hide the fact I am from Alabama.

(I'd remove the Auburn magnet from my car if I thought most people associated Auburn with the state of Alabama.)

Alabama has once again become an embarrassment to the rest of the country. 

Hey, at least you're not a bammer. No hiding it for them. A lot of idiots think Auburn is in Georgia lol. 

[AUFAN78 3,926]

After last years polling debacle, I have a hard time believing anyone would put faith in them. Insanity.

[homersapien 11,452]

1 minute ago, Bigbens42 said:

Hey, at least you're not a bammer. No hiding it for them. A lot of idiots think Auburn is in Georgia lol. 

Hush!  Let's keep it that way. 

[AUDub 11,184]

5 minutes ago, AUFAN78 said:

After last years polling debacle, I have a hard time believing anyone would put faith in them. Insanity.

The polls actually didn't do terribly. FiveThirtyEight did a great job aggregating them, were quite bullish on Trump and got lambasted because their chances for a Trump win were high compared to others like the NYT and HuffPo. 

[AUFAN78 3,926]

7 minutes ago, Bigbens42 said:

The polls actually didn't do terribly. FiveThirtyEight did a great job aggregating them, were quite bullish on Trump and got lambasted because their chances for a Trump win were high compared to others like the NYT and HuffPo. 

You can say that, but how many actually got it right? 

[AUDub 11,184]

3 minutes ago, AUFAN78 said:

You can say that, but how many actually got it right? 

You mean what percentage were within their margin of error? Turns out, practically all of them. 

https://fivethirtyeight.com/features/the-real-story-of-2016/

Another myth is that Trump’s victory represented some sort of catastrophic failure for the polls. Trump outperformed his national polls by only 1 to 2 percentage points in losing the popular vote to Clinton, making them slightly closer to the mark than they were in 2012. Meanwhile, he beat his polls by only 2 to 3 percentage points in the average swing state. 3 Certainly, there were individual pollsters that had some explaining to do, especially in Michigan, Wisconsin and Pennsylvania, where Trump beat his polls by a larger amount. But the result was not some sort of massive outlier; on the contrary, the polls were pretty much as accurate as they’d been, on average, since 1968.

[AUFAN78 3,926]

3 minutes ago, Bigbens42 said:

You mean what percentage were within their margin of error? Turns out, practically all of them. 

https://fivethirtyeight.com/features/the-real-story-of-2016/

 

 

This was not a trick question and it doesn't need a trick answer.

Let's try again. How many got it right?

[homersapien 11,452]

Simple 'argument by assertion' is a tough thing to pull off in this day and age.  

[homersapien 11,452]

Just now, AUFAN78 said:

This was not a trick question and it doesn't need a trick answer.

Let's try again. How many got it right?

Never took statistics, huh?

[jluvah 672]

32 minutes ago, homersapien said:

I just know that if Moore wins, I am so glad that I can hide the fact I am from Alabama.  (I'd remove the Auburn magnet from my car if I thought most people associated Auburn with the state of Alabama.)

Alabama has once again become an embarrassment to the rest of the country. 

Remove that magnet anyway, Hilary

[AUFAN78 3,926]

1 minute ago, homersapien said:

Never took statistics, huh?

Quantitative analysis close enough? 

Same tired game, different day. You scared to answer a simple question? If so, why? 

[homersapien 11,452]

Just now, AUFAN78 said:

Quantitative analysis close enough? 

Same tired game, different day. You scared to answer a simple question? If so, why? 

No.  Completely different subject actually.

Ben already answered it.

[AUFAN78 3,926]

Just now, homersapien said:

No.  Completely different subject actually.

Ben already answered it.

Nope and nope. Kinda early to weasel homes. Come on, you can do this. 

[homersapien 11,452]

16 minutes ago, jluvah said:

Remove that magnet anyway, Hilary

So, what's your connection to Auburn that you feel entitles you to say that newbie?

And Hillary is spelled with two "L"s.

[AUDub 11,184]

37 minutes ago, AUFAN78 said:

This was not a trick question and it doesn't need a trick answer.

Let's try again. How many got it right?

No poll is ever "right." Just like Box's statement on statistical models goes, all polls are wrong. That does not mean they are not useful. What matters is the assumptions we make from them.

But let's focus on FiveThiryEight's analysis:

Again, what matters is the assumptions we make, and in that regard FiveThiryEight did very well. NYT, HuffPo, and the like were stupid to call the race 99% Clinton, but at the end of the day they're newspapers/websites, while FiveThirtyEight are data and poll analysis experts. I think them giving Trump a 33% chance while everywhere else wrote him off completely is telling of this fact.

They even stated numerous times that if Clinton won the popular vote by only 1 or 2, Trump would win the EC most of those times. They emphasized this was a real possibility. It made a lot of people uncomfortable, and some idiot from HuffPo even felt the need to attack Silver over it on Twitter.

In short, they didn't get every state like they normal do, but they did a damn good job with a strange election. That gets lost in all the noise.

[AUFAN78 3,926]

Let me make this easy for you two. It won't hurt. Much. 

https://www.realclearpolitics.com/epolls/latest_polls/president/

[AUDub 11,184]

2 minutes ago, AUFAN78 said:

Let me make this easy for you two. It won't hurt. Much. 

https://www.realclearpolitics.com/epolls/latest_polls/president/

The answer is they didn't really get it that wrong. Nationally, the average was about C+4, and the final result was around C+2, which is well within the bounds of normal polling error.

What happened was that the polling error happened to be concentrated in EXACTLY the right places for Trump to win the electoral college while losing the popular vote by a substantial margin.

The big misses, and I emphasized this a page ago here:

1 hour ago, Bigbens42 said:

Like the rustbelt, I don't think a lot of pollsters know how to properly poll this state 

were in the rust belt, as result of nobody really knowing how to poll those states.

The justification FiveThirtyEight gave for its high Trump win chances ended up being exactly what happened. The rust belt state errors were correlated, as FiveThirtyEight said they could be. It was a close election, and there was a moderately large polling error in a few states. It was not a historic polling miss.

[TitanTiger 20,511]

17 minutes ago, AUFAN78 said:

Let me make this easy for you two. It won't hurt. Much. 

https://www.realclearpolitics.com/epolls/latest_polls/president/

Don't act like a simpleton.  National polls measure popular vote.  Most of them had Clinton winning the popular vote.  Clinton won the popular vote.  How close they got is a different story.  There is a reason that statisticians use "margin of error" and the predicted Hillary win in the national polling was within that margin.  It wasn't like they had her up +10 and it ended up +2. Even if all you took was quantitative analysis, you know better than to try and frame the question this way.  It's disingenuous and you know it.

Where people were off was in a handful of state polls.  The margins for Trump's win in most of those swing states was razor thin.  Most polls have a margin of error of 2-4%.  The margin of victory for Trump in 4 key battleground states:

Michigan:  0.3%
Wisconsin:  0.7%
Pennsylvania:  0.7%
Florida:  1.2%

Three of those four (MI, PA, WI) hadn't voted for a Republican in a presidential election since 1988 when Michigan and Pennsylvania went for Bush, Sr.  The last time Wisconsin was won by a Republican, it was Reagan in 1984.  Florida was won by Obama the previous two election cycles.  Anyone trying to make predictions in volatile and razor thin polls prior to this past election would have been totally and reasonably justified in believing Hillary would pull at least MI, PA and WI out.

So whether we're discussing national popular vote or state polls, your gnat straining, disingenuous framing of the question about "who got it right" fails when examined by anyone who engages their brain and has even a college 101 level understanding of statistics and polling.  Just stop.  This is why people don't take you more seriously.

[homersapien 11,452]

78 obviously has a problem with understanding what is meant by "margin of error".

Apparently, he thinks polls (or the results of any statistical sample) present a definitive outcome given a dichotomous choice.

 

[homersapien 11,452]

The greater question is why he thinks this is important or even meaningful. 

Care to tell us 78?

[Proud Tiger 4,261]

33 minutes ago, AUFAN78 said:

Let me make this easy for you two. It won't hurt. Much. 

https://www.realclearpolitics.com/epolls/latest_polls/president/

I guess it does hurt some. Too bad you get called names for just making a post with facts. But that seems to have become the norm here. Funny that this is the so-called Auburn "family." Sure isn't my idea of a family when people can't seem to make a post without name calling and belittling other posters because they have a different opinions. The forum rules seem to have gone down the tube.

[AUFAN78 3,926]

15 minutes ago, Bigbens42 said:

The answer is they didn't really get it that wrong. Nationally, the average was about C+4, and the final result was around C+2, which is well within the bounds of normal polling error.

What happened was that the polling error happened to be concentrated in EXACTLY the right places for Trump to win the electoral college while losing the popular vote by a substantial margin.

The big misses, and I emphasized this a page ago here:

were in the rust belt, as result of nobody really knowing how to poll those states.

The justification FiveThirtyEight gave for its high Trump win chances ended up being exactly what happened. The rust belt state errors were correlated, as FiveThirtyEight said they could be. It was a close election, and there was a moderately large polling error in a few states. It was not a historic polling miss.

You seem to enjoy these train rides. 

Nothing wrong with FiveThirtyEight and their data. I think they do a good job. Happy?

Now if we can get back to reality please. I can agree polls were close, but that wasn't the premise of my question. The vast majority of polls missed on this election. The numbers don't lie and this wasn't horseshoes.  

[AUDub 11,184]

26 minutes ago, AUFAN78 said:

You seem to enjoy these train rides. 

Nothing wrong with FiveThirtyEight and their data. I think they do a good job. Happy?

Now if we can get back to reality please. I can agree polls were close, but that wasn't the premise of my question. The vast majority of polls missed on this election. The numbers don't lie and this wasn't horseshoes.  

Your emphasis on which polls were "right" is where you're missing. All polls are wrong, but some are more wrong then others. The real question should be "were the polls more wrong than usual?" From a statistical standpoint, the answer is "not necessarily."