PA: CEPEX: Junk poll has Trump +5

[Fmr President & Senator Polnut]

https://twitter.com/NateSilver538/status/767496481054003200

Stop quoting Nate BRONZE

He made the wrong call on Trump (but again, a lot of us thought better of Republicans) and not perfect, but he does understand this issue.

It's a fake poll, or if not fake, a hideously botched one.

[GeorgiaModerate]

Soooo...it's fake?

Either that, or it's the first assignment in a freshman Stat class.

[The Other Castro]

It would be so easy for a knowledgeable person to make a realistic fake poll, and I just don't get how these people go through this effort and not even check the calculations for simple errors.

[Fmr President & Senator Polnut]

https://twitter.com/NateSilver538/status/767496481054003200

Stop quoting Nate BRONZE

He made the wrong call on Trump (but again, a lot of us thought better of Republicans) and not perfect, but he does understand this issue.

It's a fake poll, or if not fake, a hideously botched one.

Says increasingly nervous man for the 10th time this year.

What?

[GeorgiaModerate]

It would be so easy for a knowledgeable person to make a realistic fake poll, and I just don't get how these people go through this effort and not even check the calculations for simple errors.

Not only that...if someone really wanted to create a believable fake narrative [disclaimer: I don't!!!], the best way to do it would be to create a whole series of self-consistent polls that showed a gradual trend in the desired direction.  Instead of a single poll with Trump suddenly up +5 in PA, start with one that shows him down by 4 or 5 -- closer than it has been, but within the bounds of possibility.  A week later, put out another one with him down 1 or 2, then another with him up +1 or 2, and so on.  Continue the trend, and maybe even backtrack a bit at times to increase credibility.

[Xing]

Looks toats lejit, amirigt?

[Darthpi – Anti-Florida Activist]

Dammit people, if you're going to make up fake polls at least put forward the effort that Overtime Politics did and make sure the internal numbers add up correctly. Might at least give some people a nagging sense of worry in the back of their heads for a few weeks until you miss Alabama by 48 points.

[Fmr President & Senator Polnut]

Dammit people, if you're going to make up fake polls at least put forward the effort that Overtime Politics did and make sure the internal numbers add up correctly. Might at least give some people a nagging sense of worry in the back of their heads for a few weeks until you miss Alabama by 48 points.

"a "D" turns into a "B" so easily, you just got greedy"

[Senator-elect Spark]

Trump is still viable until election day, remember that folks.

[Wells]

Trump is still viable until election day, remember that folks.

Not if the only polls that show him leading are fake.

[OneJ]

Did they poll the KKK or are they just plugging in numbers?

[Seriously?]

Soooo...it's fake?

Fake? Probably not. Poorly implemented to be useless? Yes. This is a Canadian company for starters. Poling the U.S. election.

I wouldn't bet the house on Trump winning PA based on this poll.

[Wells]

Soooo...it's fake?

Fake? Probably not.

Did you even read my posts regarding this poll? The numbers are completely fabricated and don't add up.

[Alcon]

Soooo...it's fake?

Fake? Probably not. Poorly implemented to be useless? Yes. This is a Canadian company for starters. Poling the U.S. election.

Of all the reasons to reject this poll, why this one?

[Seriously?]

Soooo...it's fake?

Fake? Probably not.

Did you even read my posts regarding this poll? The numbers are completely fabricated and don't add up.

Yes. But the company exists and put this out via PR Wire. I don't think they know what they are doing, but I don't think they faked this poll by any means.

I've discounted it. It's really tom-at-o/tom-a-to as far as I am concerned.

[Darthpi – Anti-Florida Activist]

Soooo...it's fake?

Fake? Probably not.

Did you even read my posts regarding this poll? The numbers are completely fabricated and don't add up.

Yes. But the company exists and put this out via PR Wire. I don't think they know what they are doing, but I don't think they faked this poll by any means.

I've discounted it. It's really tom-at-o/tom-a-to as far as I am concerned.

Let's dispel with this fiction that they don't know what they're doing. They know exactly what they're doing.

[Maxwell]

Yeah, this is fake.

At least it is more convincing than that "leaked" PPP internal.

well, to be fair, the leaked PPP internal was meant to be funny.

[pbrower2a]

If it is a fake poll, then will the moderator kill and delete the thread?

[Fargobison]

Gravis better watch out, CEPEX might be Breitbart's new polling firm with high energy results like this.

[Desroko]

One very basic element of this poll doesn't add up.  The URL in the OP leads to this actual source: http://www.projectexecution.consulting/cepexpoll/, which says that the poll is of 1000 RV (landline numbers, English, robocall) and that "Results are statistically significant within ±7.1 percentage points, 19 times out of 20."

But the MOE should be much smaller for a sample that size.  Based on a sample of 1000 in a population of 8.5 million (PA's 2012 RV number), I get 3.1%.

FYI, population size is irrelevant to the standard error when using a sample proportion (as opposed to a sample mean). It's 3.1 for Trump and 3.0 for Clinton. 

[True Federalist (진정한 연방 주의자)]

Soooo...it's fake?

Says who!

This guy for one.

[Joe Republic]

I've entered this into the database, just to be safe.

[Speed of Sound]

I've entered this into the database, just to be safe.

And I hacked it into Silver's 538 projections for safety, as well. He'll thank me when he completely nails PA this fall.

[GeorgiaModerate]

One very basic element of this poll doesn't add up.  The URL in the OP leads to this actual source: http://www.projectexecution.consulting/cepexpoll/, which says that the poll is of 1000 RV (landline numbers, English, robocall) and that "Results are statistically significant within ±7.1 percentage points, 19 times out of 20."

But the MOE should be much smaller for a sample that size.  Based on a sample of 1000 in a population of 8.5 million (PA's 2012 RV number), I get 3.1%.

FYI, population size is irrelevant to the standard error when using a sample proportion (as opposed to a sample mean). It's 3.1 for Trump and 3.0 for Clinton.  

True, and thanks for the correction.  I shouldn't post when tired.

[IceSpear]

Let's not post fake polls. I thought we had our fill with Overtime.