By using this site, you agree to our Privacy Policy and our Terms of Use. Close
Machiavellian on 26 July 2019
irstupid said:
Machiavellian said:

I have to agree with the-pi-guy, the polls were not inaccurate.  They had Trump well within the margin of error.  Also on the electoral college level, the states Trump won were really close.  The things about polls is that they are anonymous so people have no real issue being afraid to say who they are going to vote for unlike stating it on Facebook or social media.  

The problem is not whether polls are anonymous or not, but whether you are getting the full demographics to do them. 

Let's ask some simple questions:

There is a group of voters that are wary and suspicious of polls and the government power. They would not want to do a poll for that reason. Who do you suppose those people likely voted for?

How about white collar vs blue collar. Which group is more likely to fill out an online poll? 

Margin of error was BS. I was watching the results that night. There were states that had Clinton in the polls winning by 15 points or some junk that she ends up losing or barely wining. No margin of error is that wide. 

There are hundreds of articles about how the polls were wrong. The only reason they are touting polls are correct now is because they realize their job depends on people believing their polls/analysis of the polls. 

Has this ever been any different now then any other time.  Just because its Trump doesn't mean the science or the polling has changed.  You would be highly surprised at the math models used for polling and how they take in a lot of different variances when making their report.  Trump isn't the first or the last poloraing President and he isn't outside the norm as far as how polls go.

I would like to know what polls you were watching that had Clinton at 15% that she lost in because I saw absolutely nothing like that.  In the battle states, she was only 1 to 2 percentage points in the lead and I remember in a few she should have won she was dead even which was bad and a few commented on that.

You have to remember there is a difference between the whole state and individual cities.  Also with the districting it could screw numbers as well.



Around the Network
Chrkeller on 26 July 2019

Polls were inaccurate. It is one thing to say Trump was suppose to lose within the margin of error in a state or two. But when he was suppose to lose in 7-10 states and ended up winning them all... yeah, something was wrong.  Defying the odds across the board in multiple states means their was a systematic error that was not taken into account.  Trump shockingly picked up the vast majority of 'swing states' which absolutely brings into question the accuracy of polling.  

Last edited by Chrkeller - on 26 July 2019

Chrkeller on 26 July 2019
the-pi-guy said:
Chrkeller said:

Polls were inaccurate. It is one thing to say Trump was suppose to lose within the margin of error in a state or two. But when he was suppose to lose in 7-10 states and ended up winning them all... yeah, something was wrong.  Defying the odds across the board in multiple states means their was a systematic error that was not taken into account.  Trump shockingly picked up the vast majority of 'swing states' which absolutely brings into question the accuracy of polling.  

It was 5 states according to 538's prediction, no where near 10.  4 of which were lost with a less than 2% margin.  Again most of the polls had a 3% margin of error.  

There is no "supposed to win".  That's not what polls are supposed to do.  Trump "defied" the odds in the same way I "defy" the odds by rolling a 6 with my die.  

You simply don't understand statistical modeling.  For the record I am an analytical chemist who does statistics on a regular basis.  Margin of error is plus AND minus, not plus or minus.  Meaning error is suppose to be random, not directional in a singular sense.  As an example if my specification is 8-12, with a target of 10; and I run multiple batches and come up with 11.2, 11.9, 10.9, 11.8, 11.1..  yes I am in specification.  But my error isn't random, it is systematic, which brings into question the accuracy of my model.  Being within error does not equate to the predicted mean being accurate.  But hey, I won't argue with you over this.  Believe what you want.

In practically all 11 battle states Trump scored systematically better than he was suppose to.  Being within error doesn't negate systematic error within the true mean.  



Chrkeller on 26 July 2019

"You can get random error going in one way as well, regardless of how many samples you take. "

LMFAO

I'm not wasting my time.  You don't understand statistics.  

Edit

And for clarity I have an advanced degree in analytical measurements and statistical analysis.  I have spent over 13 years doing as such, and get paid quite well doing so.  If you want to believe I don't know what I am talking about, so be it.  Can't make a horse drink and all that.  

Last edited by Chrkeller - on 26 July 2019

Chrkeller on 26 July 2019

Lol, your link literally proves my point. Thanks, much appreciated. :)



Around the Network
jason1637 on 26 July 2019

Polls were not innacurate. A lot of polls focused on popular vote which Clinton won so that's why a lot of people view them as innacurate. If you actually followed the polls in 2016 a lot of the midwest and southern states that Trump won got really close in the last week. Most polls where within the margin of errror.

Also the DOJ announced that they will allow the T-mobile and Sprint merger. It won't go through until the 13 state lawsuits are settled.
https://www.cnbc.com/2019/07/26/t-mobile-sprint-merger-approved-by-doj.html



TallSilhouette on 26 July 2019

https://youtu.be/S9VeNB5ili8



DrDoomz on 26 July 2019
the-pi-guy said:
Chrkeller said:

Lol, your link literally proves my point. Thanks, much appreciated. :)

If it proves your point, then you don't understand what I am saying.  What I am saying is that with enough randomness, you can get what doesn't look random.  None of those monkeys will systematically type out william shakespeare, but if you get enough monkey, you can get massive sets of data that make it look like one of them did.

If you had an infinite number of monkeys or random character generators, you'd get results like this:

Monkey A: AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA...

Monkey Z: ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ...

Monkey AZ: AXDXDXFKYAGKJFHLASHDFLKHDDDDDDDASLIDJF:ALDJFADF::ADJF:LJ

And you'd get everything in between, some would show randomness in one direction, most would show randomness in both directions of an average.  

That doesn't mean random character generator A is picking the letter A systematically, it is picking exactly as Monkey AZ does.

I think what Chrkeller is saying is that while something is possible, it is not probable (correct me if I am wrong). Stats landing towards one side of the margin of error multiple times consecutively aka. landing in the plus and not minus) or within a fixed pattern has such a low probability happening that statisticians would actually conclude a systematic error rather than it being random (essentially while you have the same chance as landing heads as tails and it is possible to land heads 7 times, it landing 7 times in a row has such a low possibility of happening that one would wonder if the game was somehow rigged).

But unlike coin flipping, where it is hard to find any other possible way of corrupting the data, in polling, there are so many things that can (such as sampling bias) that if such a low probability event happened, one would lean towards a systematic error rather than a freak occurrence.

Not sure if you guys covered this (and somehow still didn't come to an understanding somehow) since I didn't read your entire exchange. So excuse me if this has already been covered.



Chrkeller on 26 July 2019
DrDoomz said:
the-pi-guy said:

If it proves your point, then you don't understand what I am saying.  What I am saying is that with enough randomness, you can get what doesn't look random.  None of those monkeys will systematically type out william shakespeare, but if you get enough monkey, you can get massive sets of data that make it look like one of them did.

If you had an infinite number of monkeys or random character generators, you'd get results like this:

Monkey A: AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA...

Monkey Z: ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ...

Monkey AZ: AXDXDXFKYAGKJFHLASHDFLKHDDDDDDDASLIDJF:ALDJFADF::ADJF:LJ

And you'd get everything in between, some would show randomness in one direction, most would show randomness in both directions of an average.  

That doesn't mean random character generator A is picking the letter A systematically, it is picking exactly as Monkey AZ does.

I think what Chrkeller is saying is that while something is possible, it is not probable (correct me if I am wrong). Stats landing towards one side of the margin of error multiple times consecutively aka. landing in the plus and not minus) or within a fixed pattern has such a low probability happening that statisticians would actually conclude a systematic error rather than it being random (essentially while you have the same chance as landing heads as tails and it is possible to land heads 7 times, it landing 7 times in a row has such a low possibility of happening that one would wonder if the game was somehow rigged).

But unlike coin flipping, where it is hard to find any other possible way of corrupting the data, in polling, there are so many things that can (such as sampling bias) that if such a low probability event happened, one would lean towards a systematic error rather than a freak occurrence.

Not sure if you guys covered this (and somehow still didn't come to an understanding somehow) since I didn't read your entire exchange. So excuse me if this has already been covered.

This.



sundin13 on 26 July 2019

I'd just like to add this insight from 538:

Overall, while there was a fairly significant lean in the 2016 election cycle, if you take another step back, you see that the pattern tends to fade away. While it is fairly common for leans to occur, they tend to unpredictably swing back and forth. There isn't really evidence that there are huge issues with polling that weren't present in previous years.

The biggest issue with polling is still people not understanding polls.