Be a statistician and think about it for a second

Be a statistician and think about it for a second...

In a study when you have a variable amount of people involved, when do you have more accuracy? When you have a higher number of people involved in the study? Or when you have a lower amount of people involved in the study?

These poll graphs on all the websites with this perfect pattern have got so many people fooled it's amazing.

When the two lines approach each other after the first bubble, we were approaching the end of the conventions and polling numbers were high.

When the two lines approach each other after the second bubble, we were approaching the first presidential debate.

Notice how on 11/01/2016, when 538 people recognized how stupid their model is, they changed it momentarily perhaps so it would look a bit more volatile...but I caught them and sent them a message asking why it changed and they changed it back. Too bad, I have a screenshot.

And so now we are approaching the actual election...well it's no secret what is going to happen.

Other urls found in this thread:

uniteforsight.org/global-health-university/importance-of-quality-sample-size
twitter.com/SFWRedditGifs

You have more accuracy when your sample better reflects the population.

large, representative, non-biased

Are you just saying that 538 is activly modifying their parameters for their polling?

Just give me a simplistic run down of what youre trying to put across here

You're pretending like they don't just make up responses. There's no oversight on private polls like this, they can lie their asses off. Not even any laws against it.

your chart on 11/1 shows now-cast, the others are polls-only

I agree completely. This is a bad model.

That's not necessarily what I'm putting across, though that is true. They can change their models however they feel both best represents their data as well as best represents them.

This all falls under the 538 "Chance of Winning" model, which should be reflective of their polling data.

But it would cost them a lot of money

If they are really as wrong as you believe they are, then what are they going to do next election when their brands are utterly ruined?

My take is that if they really are rigging the polls, it's only so that they more accurately reflect the rigged results of the election

If the sample follows the normal distribution then a larger size is better. If the sample is biased or skewed in any way then a larger sample will not help.

All of these kinds of polls are pretty suckish.

When was the last time they were totally wrong? Reagan?

The idea behind a statistical analysis of a pattern in a trend like this coupled with details about the context of the changes, is that it helps you predict what the 'unseen' data is.

If there is a larger amount of people involved in the poll, it will be more accurate.
The direction the trend changes to during the period of time there are larger polling numbers represents a fraction of the direction it would have when all results are in.

I think it's quite odd how whenever Trump gets a spike, Hillary gets a spike too or stays neutral. I wonder if you could superimpose an undecided voter graph on this, because that's kind of the only place any 'total inflation' can come from. For example, right now we are at 93 ish % of total voters, and that seems to be fluctuating quite a bit, but only goes up when Trumps numbers go up. Odd.

It looks that way because Johnson has such low polling numbers his trend line has no effect on Trump's or Hillary's.

Since Johnson's % has been 0.1% the entire time, you could have either Hillary's or Trump's trend line on there and know what the other one is. If Johnson was doing better, that wouldn't be the case.

H(T) = 100 - 0.1 - T
T(H) = 100 - 0.1 - H

>what are they going to do next election when their brands are utterly ruined?

You mean like how the media's brands are ruined now too?

You don't get it, they are literally being paid to create fake polls by Soros et al.

They don't make money being accurate, they make money to put our propaganda polls to show whatever their masters want them to show.

Right but like for example immediately after the FBI dropped their bomb, Trump spiked and Hillary spiked before she barely fell. I don't believe the polls at all, but this seems like evidence of tampering to me. I also find it extremely hard to believe that Clinton fell less % points in her scandal than Trump did in his pussy grabbing one.

You have maximum accuracy when the sample is representative. A bigger sample doesn't necessarily reflect the subject better, no?

But for a student who is learning statistics, can you explain how 538's model work?

honestly, i have an explanation for this.
It's clear that there is a periodic pattern at play here, how is this possible?

Well some polls are biased towards clinton some towards trump, we all know that. Now, each poll comes out with its own periodicity, and imo that's the reason that chart varies periodically.

The periods where trump is close to clinton are the moments where mostly polls tht favor trump come out, and vice versa. All polls have their periodicity.

Anyway, trump will win, i already bet a lot on him and I'm sure he will, the polls are screwed cause voter turnout will be completely different

Guys...

This is going to sound crazy but I think their polls might be a load of bullshit.

I disagree. I think it looks periodic because they are skewing clintons numbers.

It should look like this

XXXXXXXXX

but it looks like a sideways hourglass instead.

How the fuck is it possible that Trump has NEVER been ahead of Clinton. It's not fucking possible. That alone gives away the fact that Clinton is oversampled to STAY AHEAD

the big, real issue that most people are painfully ignorant to is that democrat voting was down in the primaries, and republican voting & registration & independent voting was way, way up.

the polls being used to support clinton's chances at victory are oversampled in favor of democrats due to historic voting patterns. they do not consider the trends from the primaries.

it's a "yuge" mistake.

When FBI dropped their bomb, the trend was already moving in the direction it is. The FBI dropping the bomb absolutely did speed up the trend. Trump's pussy grabbing one was big...but I think the general population understands the difference in importance about a guy trying to impress his mates and maybe predecessors, and war with Russia.

That's not always true, especially not with people. Consider medical studies and psychological studies as a comparable example. This page outlines it pretty well, it has more information on it than I care to regurgitate right now: uniteforsight.org/global-health-university/importance-of-quality-sample-size

I'm not sure what you are getting at with periodicity desu. But yeah, there are many bias polls....just like the media.

If the polls were correct, they would be nearly flat lines nearly the whole time. It shouldn't fluctuate more than 5%. Not many people flip flop their candidate choice.

Let me elaborate

If Trumps number go down pretty much in line with whenevrr Hillarys numbers go up, then we should expect to see the exact same trend when Clinton goes down and Trump goes up, but we don't!

What we see is Clinton going up or slightly down, and the total number of decided voters going up.

Sorry but this is fucking horseshit.

See >>>

There is no # of decided voters represented in this graph or these polls

This is actually quite interesting, I'm doing research and publishing a paper aalyzing these elections, and we've built a model that tries to predict movements in the polls by reading thousands of rss news feeds. What we see is that the movement in trump's poll values are quite correlated with news and can be predicted better, while the movents in clinton's polls are much more constant and less correlated with news.

This may be either due to clinton voters not giving a shit about news, or to some kind of poll rigging i guess. By naked eye it seems as if clinton and trump polls move exactly specularly, but that's not true, or our algorithm would be able to predict movements in clintons polls just as well as movements in trump's polls. In fact trump polls can have much bigger swings than clinton polls, and i'm not really sure why

Interesting read. Thanks.

>I'm not sure what you are getting at with periodicity desu.

so each pollster has a sample that is skewed towards one candidate. Each pollster makes polls come out at their own periodicity (some daily, some every other day, or every week). These polling averages average the latest polls, so there is bound to be some periodicity effects here, do to polls being released in a periodic fashion. Now im not sure how often each pollster updates their polls so i havent really looked into this, just an idea.