I find it interesting how yesterday's victims are almost always today's aggressors. The transfiguration that the Jewish people suffered between WWII and today's Israel policies is impressive. It's almost impossible to recognize this people today. And it took them almost no time to go from victim to aggressor in the international stage. Which really should tell a lot about them.
But because I am saying this I must be Muslim or antisemitic.
Reminds me of one of Trump's argument that most resounded with me. The horrible, horrible idea of arming the Curds and providing the Syrian rebels with logistic support. When he told Hillary that we don't know who these people are, I almost jumped off my chair. F... hell! Finally!
It's not just that in the whole history of American interventionism almost everyone USA armed or helped financially during armed conflict in the middle-east ended up resulting in the loss of American lives not 20 years later at the hands of those they helped. It's also not only because of the fact that USA was one (the biggest) contributor to the rebirth of religious terrorism in the past 50 years. It is also the fact that you don't arm an entire combat force founded on radical religious and cultural principles and expect that to turn out well. And it is particularly stupid when you do that for the sake of the resolution of a single local conflict, without a care in the world for the fact you know you won't be able to remove those arms once it is over.
Last edited by Mario F.; 10-11-2016 at 12:43 PM.
Originally Posted by brewbuck:
Reimplementing a large system in another language to get a 25% performance boost is nonsense. It would be cheaper to just get a computer which is 25% faster.
Agree on all points. The quote "yesterday's victims are almost always today's aggressors" is going to stay with me a long time, made me think a bit.
In general, I think the US should keep its nose out of other peoples' problems that are clearly only their problems. Talking armed forces-wise, not humanitarian aid etc. Realistically, will there ever be peace in the middle east?
One of these days, I ought to read some of the agreements we have with other countries to see how feasible this is. I do dislike that we engage in a lot of conflicts.
It's my understanding that a lot of modern military engagements aren't strictly war, but the US goes to NATO or the UN and then we form a coalition and attack. So toning down military conflict seems to be about shaping opinion, and saying we don't care for pre-emptive attacking or arming Israel. If I'm correct then any real change is going to be slow moving because America's democracy is pretty sick at the moment. I think a good letter writing campaign would get the ball rolling though.
"...a computer is a stupid machine with the ability to do incredibly smart things, while computer programmers are smart people with the ability to do incredibly stupid things. They are,in short, a perfect match.." Bill Bryson
>> What poll are you referring to? CNN's 500-person poll that oversampled Democrats?
>> Because independent polls not undertaken by the establishment overwhelmingly show otherwise:
>> Thing to remember is that we have 320 million people, and these 500-2000 sample sizes are crap. Polls are useless, just wait for the big tamale.
I know this is a week old, but I'm surprised nobody responded to these messages. Regardless of how you feel about the candidates, at least educate yourselves on how polling works before making these types of comments.
First, to Yarin. The "not undertaken by the establishment" polls that appear to be in that picture are completely different than the polls that showed that Clinton "won" the debates. The polls in those pictures are merely surveys open to anybody who decides to click and answer them. This makes the population who respond not representative of the country as a whole. If the website doing the survey has an audience that prefers one candidate over another, then you will likely get an outcome in the survey favoring that candidate. If one candidate's supporters are more enthusiastic, especially about the outcome of these surveys, then you also might get an outcome in the survey favoring that candidate.
This is different than the scientific polls that showed Clinton "won" the debates. Scientific polls attempt to survey a sample of the country (or the subset of residents that might vote in the election). While there is certainly some margin of error in doing this kind of poll, it can generally be relied upon to give a rough estimate of how the entire population feels. If you look at several different independent scientific polls, you can get a pretty good sense of who people (or voters) feel "won" the debates.
In the last U.S. Presidential election, many Romney supporters didn't think the polls (which had a fairly strong consensus that Obama was up by about 2 points) were accurate. They thought that Romney's supporters were more enthusiastic, and that the polls oversampled Democrats. It turns out, the polls were pretty accurate and could be used to accurately predict the outcome in all 50 states. In fact, the polls underestimated Obama's support, and he won by about 4 points nationally.
The lesson is, don't ignore the "mainstream media" and "establishment" polls because you're biased against them, and definitely don't look at unscientific online surveys as evidence of anything. If you average the results of a few national polls, chances are you'll be within a couple points of the actual state of the race.
And Epy, I'm not sure why you think a sample size of 2000 out of the 225 million eligible voters is crap, especially when it is weighted scientifically. As I mentioned before, poll aggregators predicted the national vote to within a few points and predicted each state outcome last election and the election before. I mean, sure, one individual poll might be off by 10 points but there's no reason to wait for the big tamale to understand that (a) Clinton currently has a big lead and (b) her lead increased after the first and second debates.
Another terrific sample size. How many presidential elections have we had? Historically, some elections have had huge changes over the course of months, and some have been as bad as 5% off between the election results and the last polls. Here's another great sample size: I only looked at Gallup's historical dataOriginally Posted by Daved
Election Polls -- Accuracy Record in Presidential Elections | Gallup Historical Trends
Gallup Presidential Election Trial-Heat Trends, 1936-2004
Because statistically, for a single poll, the best you can get assuming everything went perfectly (perfectly representative sample) is +/- 3%. That number increases as the population proportion gets farther away from 50%.Originally Posted by Daved
Assumptions: average sample size of polls = 1000, election is tied 50-50, 95% confidence.
p = population proportion
n = sample size
Standard error = sqrt(p(1-p)/n) = sqrt(0.5(1-0.5)/1000) = 1.58%
Multiplied by t-value for 95% confidence (1.96) = 3.1%
Confidence interval is then p +/- 3.1%
Play with that math if p <> 0.5; confidence interval gets bigger / worsens.
Basically, can only begin to suspect that the poll is meaningful if the spread is >6.2%. But that's absolute best case scenario.
Now if you take a bunch of different polls to make an average and do a CI for that...
Taking data from: Latest Election Polls 2016 - The New York Times
results:Code:clc clear alpha=0.05; % 95% CI n=10; % sample size ts = [tinv(alpha/2,n) tinv(1-alpha/2,n)]; % critical t values % hillary x=[.47 .42 .42 .40 .45 .5 .43 .51 .51 .50]; CI=mean(x) + ts*std(x)/sqrt(n) % trump x=[.4 .42 .38 .41 .39 .41 .39 .36 .46 .38]; CI=mean(x) + ts*std(x)/sqrt(n)
So today, you can say that Hillary will probably win because the CIs do not overlap. But if you change the confidence level to 99%, you can't say that.Code:CI = 0.43121 0.49079 CI = 0.38063 0.41937 >>
Moving on though, look at that same link above and look at the percentages over time...point is, yeah, if we had the election tomorrow, Hillary will probably win, but it's still 3 weeks away so these polls will become meaningless.
But the larger point is not who is going to win, the point is that you need huge sample sizes to really get accurate and meaningful, and that you have to wait for the polls that are very close to the election date.
Again, keep in mind this is all assuming that everything is perfectly unbiased. Lols.
Aside: who has the time or desire to actually answer phone surveys? Or any form of that crap? That's what you call sample bias.
You call CNN's poll scientific, but you never once mention what methods they employ to ensure the accuracy of their polls. I'm sure you're not just slapping the "science" label on something just because you agree with it... right?
Most of the western media has a sickeningly evident bias. Regardless how you feel about the candidates. You would think CNN would be reporting on the leaked emails and Veritas tapes that amount to admissions to fraud, threatening the integrity of your democracy. Instead, they're too busy interviewing athletes on what they think about Trump's use of the term "locker room talk", and too busy attacking him for not objecting to a talk show host using a vulgar term for his daughter.
I do trust the judgment of the people who sat in the room with them during the debate, who applauded and laughed for Trump, and actually booed Clinton, over a poll conducted by an organization who's bias is so bloody obvious that it's disgusting.
>> I'm sure you're not just slapping the "science" label on something just because you agree with it... right?
Of course not. There is a fundamental difference between how scientific polls are administered and how online surveys are administered. It doesn't matter which side the results look good for. (In fact, I kept putting "won" in quotations because the idea of winning a Presidential debate is kind of dumb in the first place. I don't care who "wins" these debates.) If you can't acknowledge that fundamental difference, then you're basically burying your head in the sand. (And again, you wouldn't be the first. Romney supporters did this in 2012. Kerry supporters did this in 2004. Don't be one of those people.)
>> I do trust the judgment of the people who sat in the room with them during the debate
Sure, you can do that. You can also ignore what others say and make a judgment yourself. What you should avoid doing, though, is thinking that the country agrees with you because some unscientific online surveys say that is true when the vast majority of actual polls say otherwise.
Also, note that the CNN poll has shown very little house effect historically (source). People probably claimed it was biased in 2012 and 2008 and so on, but they ended up being wrong.
An important part of holding any belief (especially one that is being challenged) is to consider how you could be proven wrong. What experiment would you do to test whether your idea was wrong? In this case, if Clinton wins on Nov 8 and her national vote margin is within 3 points either way of the RCP, 538 or PEC poll aggregations, will that make you reconsider? Note that I expect the final vote could be off by more than that, so by itself a big miss wouldn't shake my belief in the relative accuracy of aggregations of scientific polls, but if they were off by 8-10 points it might.
>> But if you change the confidence level to 99%, you can't say that.
Ok, so what you were saying is that polls can't tell us the winner with greater than 99% certainty. I agree with that. Or really that polls are fallible. I agree with that, especially individual polls. It's just when you said polls are useless that I disagreed with. They are incredibly useful and relatively accurate in explaining the state of a national race.
I do think most touted polls, including CNN's, do/will accurately reflect the outcome of the election. However, this is distinctly different from public opinion.
Any poll conducted by a neutral party, on a larger sample, with an open sampling process and data to prove accuracy would suffice.
It's somewhere in between 95 and 99%, and that would be the winner if the election was held tomorrow. If all that is unbiased.
Polls are good iff: unbiased, have large sample sizes (IMO need >5000, more the better), and if many are used together as a whole to draw conclusions (you need many trials, not just a lot of samples). Many good polls, averaged together, can tell the state of the race at that point in time. Look at the link from NY Times, it shows how the national averages have swung in the past couple of months. There are many points in time that the CIs would overlap because the spread isn't big enough.
But even so, again, not 100% of people do surveys. I know I don't and know many others don't. Don't have time for it during the day. That's part of that last poll vs. actual election difference, among other things.
>> Any poll conducted by a neutral party, on a larger sample, with an open sampling process and data to prove accuracy would suffice.
Can you name what you'd consider to be a neutral party? There are dozens of people doing polls of the election (see my previous link). Which of those do you consider to be a neutral? There are usually a couple different pollsters that do "who won the debate" polls. I'm guessing most of them will show Clinton to be the "winner" of the third debate based on what I've seen so far, so maybe you'd say that none are neutral. But ideally you would name what you consider to be neutral polls ahead of time, then we'd see what those polls show. That would be a proper test for your theory.
Related but not on current topic, the more I see of this ........ election the more I want to leave the country.