|
Read the rules in the OP before posting, please.In order to ensure that this thread continues to meet TL standards and follows the proper guidelines, we will be enforcing the rules in the OP more strictly. Be sure to give them a re-read to refresh your memory! The vast majority of you are contributing in a healthy way, keep it up! NOTE: When providing a source, explain why you feel it is relevant and what purpose it adds to the discussion if it's not obvious. Also take note that unsubstantiated tweets/posts meant only to rekindle old arguments can result in a mod action. |
On March 01 2017 01:04 SoSexy wrote:Show nested quote +On March 01 2017 01:02 Plansix wrote:On March 01 2017 00:59 SoSexy wrote:On March 01 2017 00:57 Acrofales wrote:On March 01 2017 00:47 SoSexy wrote:On March 01 2017 00:45 Mohdoo wrote:On March 01 2017 00:43 SoSexy wrote:On March 01 2017 00:41 Mohdoo wrote:On March 01 2017 00:37 SoSexy wrote:On March 01 2017 00:23 Mohdoo wrote:
[quote]
What is your statistics background? Are you aware that things with a 1% chance of happening, do indeed happen every day? Its not like once something makes it past 60%, it is a guarantee. Indeed. But I expected one of the biggest american newspaper to have access to better resources than me at home. But that's the thing. A PhD statistician can give something an 86% chance of happening, and it does not mean it will happen. A legion of statisticians can tirelessly work to give the most accurate probability as possible, but it will always be a probability. There is no shame in something with an 86% chance of happening, not happening. Rolling snake eyes has a 2.7% chance of happening, but it happens. Your defense makes no sense. Applying this logic, one could defend basically everything because there is a 'chance of it happening'. Decisions do not work in this way. In short, they could have been more careful and just put a 60%. I would have been happier with that. How do you think probabilities are calculated? What do you think the process looks like? In a presidential election? Probably a mix of past results and recent polls. Yes. And they are then combined using statistical techniques to calculate the likelihood for that event. If that process says there's an 86% chance of it happening, you then think they should post "60% chance" because they "need to be more careful"? Wut? No, it means their polls (and instruments to understand what the american population believes) are shit-tier. Polls can be wrong. They are not perfect tool. And you can’t tell if the poll was wrong until after the thing you were polling about happens. Also because polls are published, they impact future polls. People may not respond or could change their mind in the last couple of days. Of course. But i'm sure you would agree that if the final result is (stupid example) 6-1, the guy who predicted 5-1 is closer to reality than the guy who predicted 1-6?
You realize these models and stats have been used for longer than just a single year, right?
|
On March 01 2017 01:07 SoSexy wrote:
Acrofales how many strawmen do you wanna use? Just answer this question: did the NYT do a good job in predicting the 2016 US elections? Yes or no. The rest are just cheap insults that you like to throw to strengthen your ego. I argue that it did terrible.
You could argue modern polling agencies should have been able to predict sharp changes in our society, but I think sometimes the world is a crazy place. But I must emphasize that very smart people used very good methods to deliver probabilities based on data. But if the underlining assumption that the data used is proper data for predicting, is not valid, the whole thing falls apart. That's what happened. The data wasn't meaningful anymore because society changed. All the methods were good, we just didn't have the right data.
|
On March 01 2017 01:12 Acrofales wrote:Show nested quote +On March 01 2017 01:07 SoSexy wrote:
Acrofales how many strawmen do you wanna use? Just answer this question: did the NYT do a good job in predicting the 2016 US elections? Yes or no. The rest is just cheap insults that you like to throw to strenghten your ego. They weren't predicting. They gave a statistical estimate. If you throw the NYT die, on average it will land Trump 15 out of a 100 times. We just happen to live in one of those 15 worlds. Or maybe the underlying data was wrong. That's also a possibility, and the actual a priori chances of Trump winning were far higher than 85% (and 538 used a different model in which they only estimated a 71% chance of a Clinton victory). But just because the less likely outcome happened doesn't automatically mean the a priori model was wrong. Immediately jumping to that conclusion is just a horrid understanding of probabilities.
Eh. This is one of the things I am defending.
|
United Kingdom13775 Posts
Wrote my long post about this earlier. The problem with the NYT and such is that they were just straight up using ineffective predictive techniques and likely had their desired result in mind.
And the undecideds swung strongly for Trump in the very end.
|
On March 01 2017 00:01 {CC}StealthBlue wrote:
"We gathered an unbelievable amount of intelligence that will prevent the potential deaths or attacks on American soil," said Spicer.
Hey, he finally said something I can agree with. The amount of intelligence they gathered really does seem unbelievable.
|
On March 01 2017 01:04 SoSexy wrote:Show nested quote +On March 01 2017 01:02 Plansix wrote:On March 01 2017 00:59 SoSexy wrote:On March 01 2017 00:57 Acrofales wrote:On March 01 2017 00:47 SoSexy wrote:On March 01 2017 00:45 Mohdoo wrote:On March 01 2017 00:43 SoSexy wrote:On March 01 2017 00:41 Mohdoo wrote:On March 01 2017 00:37 SoSexy wrote:On March 01 2017 00:23 Mohdoo wrote:
[quote]
What is your statistics background? Are you aware that things with a 1% chance of happening, do indeed happen every day? Its not like once something makes it past 60%, it is a guarantee. Indeed. But I expected one of the biggest american newspaper to have access to better resources than me at home. But that's the thing. A PhD statistician can give something an 86% chance of happening, and it does not mean it will happen. A legion of statisticians can tirelessly work to give the most accurate probability as possible, but it will always be a probability. There is no shame in something with an 86% chance of happening, not happening. Rolling snake eyes has a 2.7% chance of happening, but it happens. Your defense makes no sense. Applying this logic, one could defend basically everything because there is a 'chance of it happening'. Decisions do not work in this way. In short, they could have been more careful and just put a 60%. I would have been happier with that. How do you think probabilities are calculated? What do you think the process looks like? In a presidential election? Probably a mix of past results and recent polls. Yes. And they are then combined using statistical techniques to calculate the likelihood for that event. If that process says there's an 86% chance of it happening, you then think they should post "60% chance" because they "need to be more careful"? Wut? No, it means their polls (and instruments to understand what the american population believes) are shit-tier. Polls can be wrong. They are not perfect tool. And you can’t tell if the poll was wrong until after the thing you were polling about happens. Also because polls are published, they impact future polls. People may not respond or could change their mind in the last couple of days. Of course. But i'm sure you would agree that if the final result is (stupid example) 6-1, the guy who predicted 5-1 is closer to reality than the guy who predicted 1-6?
That's not the same as the guy who predicted 5-1 having the better model though.
If I say in 10 coin flips they're going to all come up heads and you say 5 heads and 5 tails, even if they come up all heads your model was better. Don't confuse luck with good modeling.
in NYT's case it was a bit too optimistic (I think 538's general criticism of other outlets is they assumed the states would behave more independently than they would which seems about right), but any model that put Trump ahead %-wise would have just been lucky rather than good. Well unless they *really* had some extra data to back their model that no one else did.
|
On March 01 2017 01:14 SoSexy wrote:Show nested quote +On March 01 2017 01:12 Acrofales wrote:On March 01 2017 01:07 SoSexy wrote:
Acrofales how many strawmen do you wanna use? Just answer this question: did the NYT do a good job in predicting the 2016 US elections? Yes or no. The rest is just cheap insults that you like to throw to strenghten your ego. They weren't predicting. They gave a statistical estimate. If you throw the NYT die, on average it will land Trump 15 out of a 100 times. We just happen to live in one of those 15 worlds. Or maybe the underlying data was wrong. That's also a possibility, and the actual a priori chances of Trump winning were far higher than 85% (and 538 used a different model in which they only estimated a 71% chance of a Clinton victory). But just because the less likely outcome happened doesn't automatically mean the a priori model was wrong. Immediately jumping to that conclusion is just a horrid understanding of probabilities. Eh. This is one of the things I am defending.
So if these same methods worked very well in 2008 and 2012, what do you think was different in 2016?
|
On March 01 2017 01:14 SoSexy wrote:Show nested quote +On March 01 2017 01:12 Acrofales wrote:On March 01 2017 01:07 SoSexy wrote:
Acrofales how many strawmen do you wanna use? Just answer this question: did the NYT do a good job in predicting the 2016 US elections? Yes or no. The rest is just cheap insults that you like to throw to strenghten your ego. They weren't predicting. They gave a statistical estimate. If you throw the NYT die, on average it will land Trump 15 out of a 100 times. We just happen to live in one of those 15 worlds. Or maybe the underlying data was wrong. That's also a possibility, and the actual a priori chances of Trump winning were far higher than 85% (and 538 used a different model in which they only estimated a 71% chance of a Clinton victory). But just because the less likely outcome happened doesn't automatically mean the a priori model was wrong. Immediately jumping to that conclusion is just a horrid understanding of probabilities. Eh. This is one of the things I am defending.
Yes, but that doesn’t mean they did a bad job. Only that the data they gathered was inaccurate or was missing information. That doesn’t mean the information has zero value or something cannot be learned from it.
On March 01 2017 01:20 Mohdoo wrote:Show nested quote +On March 01 2017 01:14 SoSexy wrote:On March 01 2017 01:12 Acrofales wrote:On March 01 2017 01:07 SoSexy wrote:
Acrofales how many strawmen do you wanna use? Just answer this question: did the NYT do a good job in predicting the 2016 US elections? Yes or no. The rest is just cheap insults that you like to throw to strenghten your ego. They weren't predicting. They gave a statistical estimate. If you throw the NYT die, on average it will land Trump 15 out of a 100 times. We just happen to live in one of those 15 worlds. Or maybe the underlying data was wrong. That's also a possibility, and the actual a priori chances of Trump winning were far higher than 85% (and 538 used a different model in which they only estimated a 71% chance of a Clinton victory). But just because the less likely outcome happened doesn't automatically mean the a priori model was wrong. Immediately jumping to that conclusion is just a horrid understanding of probabilities. Eh. This is one of the things I am defending. So if these same methods worked very well in 2008 and 2012, what do you think was different in 2016?
This is the important question. What changed and why did it change.
|
On March 01 2017 01:14 SoSexy wrote:Show nested quote +On March 01 2017 01:12 Acrofales wrote:On March 01 2017 01:07 SoSexy wrote:
Acrofales how many strawmen do you wanna use? Just answer this question: did the NYT do a good job in predicting the 2016 US elections? Yes or no. The rest is just cheap insults that you like to throw to strenghten your ego. They weren't predicting. They gave a statistical estimate. If you throw the NYT die, on average it will land Trump 15 out of a 100 times. We just happen to live in one of those 15 worlds. Or maybe the underlying data was wrong. That's also a possibility, and the actual a priori chances of Trump winning were far higher than 85% (and 538 used a different model in which they only estimated a 71% chance of a Clinton victory). But just because the less likely outcome happened doesn't automatically mean the a priori model was wrong. Immediately jumping to that conclusion is just a horrid understanding of probabilities. Eh. This is one of the things I am defending.
All the NYT said was: given the data we have from <insert polls used> and the statistical model we used (which has been criticized, also before the election, but is still defensible regardless) we calculate an 85% chance that Clinton will win. Where is that sloppy or wrong reporting?
On February 28 2017 23:58 SoSexy wrote:
Never forget that '85% chances of Clinton winning' on the NYT website before the counting of votes started.
Insinuates there is something inherently wrong with putting up a win probability before an election? I mean... you could make a moral argument that trying to estimate the elections in such a way is wrong. But I don't think you were trying to do that.
|
United Kingdom13775 Posts
More undecided voters. Also both candidates were deeply despised.
Let's not make it too complicated.
|
On March 01 2017 01:22 LegalLord wrote:
More undecided voters. Also both candidates were deeply despised.
Let's not make it too complicated.
Also so much discussion about polling data may have caused some demographics to stop responding to polls all together.
|
Everyone seems to be ignoring the fact that Clinton won the popular vote by almost 3 million votes. A lot of polls had that right. If you told everyone she'd get that margin before the race, 99% of people would assume she won the presidency.
If there is a flaw in the system used by the NYT and others it is that too much focus was put on national polls. Whereas in reality we know your votes don't count unless you live in a swing state.
|
On March 01 2017 01:22 LegalLord wrote:
More undecided voters. Also both candidates were deeply despised.
Let's not make it too complicated.
An electoral/popular split also seems appropriately hard to account for. For example in this case several states had a margin of victory of <1%.
Then on top of that you had a lot of late breaking news and polls seem slower to adjust to news than people's opinions (you don't want your model to only use the most recent poll).
|
United Kingdom13775 Posts
On March 01 2017 01:30 On_Slaught wrote:
Everyone seems to be ignoring the fact that Clinton won the popular vote by almost 3 million votes. A lot of polls had that right. If you told everyone she'd get that margin before the race, 99% of people would assume she won the presidency.
If there is a flaw in the system used by the NYT and others it is that too much focus was put on national polls. Whereas in reality we know your votes don't count unless you live in a swing state.
Turnout turned out higher in the places where people were most strongly in favor of one candidate or another. Trump got a massive boost in rural America, Clinton got a massive boost in NYC, LA, and SF.
|
On March 01 2017 01:38 LegalLord wrote:Show nested quote +On March 01 2017 01:30 On_Slaught wrote:
Everyone seems to be ignoring the fact that Clinton won the popular vote by almost 3 million votes. A lot of polls had that right. If you told everyone she'd get that margin before the race, 99% of people would assume she won the presidency.
If there is a flaw in the system used by the NYT and others it is that too much focus was put on national polls. Whereas in reality we know your votes don't count unless you live in a swing state. Turnout turned out higher in the places where people were most strongly in favor of one candidate or another. Trump got a massive boost in rural America, Clinton got a massive boost in NYC, LA, and SF.
That's a pretty big simplification, there's a fair number of non-coastal places where the margin was smaller than it was in 2012. Arizona, Georgia, and Texas for example. A few rural places just barely made it over the line (Michigan, Wisconsin, and Pennsylvania) even if others tipped heavily (Ohio and Iowa).
They also lost share in a lot of other states in 2016, but it was lost to 3rd parties. 14 of the states in '16 went to someone who didn't have a majority of the vote compared to 0 of the states in '12.
|
People are ignoring the popular vote beacuse the popular vote doesn't matter.
|
That's nonsense, the popular vote has literally always played a role in determining the extent to which the executive is considered under a mandate of the people relative to its agenda directives.
Edit: Ok, so that may have not been the case in 1824 for weird reasons, but the point remains 
|
How did this thread go back to why or by how much the polls were wrong? Is it not more pertinent to discuss this absurd accusation of Obama now orchestrating the protests around the US/Globe in response to this moron's policies? Or are you guys just sick of that too now since he seems to spout some random BS every other day which takes head line news world wide by storm? I suppose polling is easier to understand than Trump logic..
|
On March 01 2017 02:11 MyTHicaL wrote:
How did this thread go back to why or by how much the polls were wrong? Is it not more pertinent to discuss this absurd accusation of Obama now orchestrating the protests around the US/Globe in response to this moron's policies? Or are you guys just sick of that too now since he seems to spout some random BS every other day which takes head line news world wide by storm? I suppose polling is easier to understand than Trump logic..
Just because you missed your check doesn't mean there's a reason to get all mad at everyone else who did get their protest check.
|
On March 01 2017 02:10 farvacola wrote:That's nonsense, the popular vote has literally always played a role in determining the extent to which the executive is considered under a mandate of the people relative to its agenda directives. Edit: Ok, so that may have not been the case in 1824 for weird reasons, but the point remains
So it has plated a role in public perception of a position that doesn't have to really worry about public perception for almost three years.
The popular vote meaning anything more then a nonsense press issue in the current system is just a petty and pointless argument against executive power.
|
|
|
|
|
|
EPT
