PILOT SPIN
Spin Zone => Spin Zone => Topic started by: JeffDG on June 28, 2016, 11:41:15 AM
-
OK, so all the polling lately has been making me think: How good are these polls? This isn't an anti-Trump screed, just some analysis. Feel free to review and critique my math if you like, it's been a bit since I did a stats class!
https://goo.gl/kRyzL7
So, I started with the RCP Average (http://www.realclearpolitics.com/epolls/2016/president/us/general_election_trump_vs_clinton-5491.html). It shows Clinton +6.8. I dropped off the bottom Bloomberg one because I just can't find out who did the poll, bringing the average to +6.2.
Next up, I went to the fivethirtyeight.com Pollster rankings. He has a very detailed analysis of how he derives the bias rankings of the polls (http://projects.fivethirtyeight.com/pollster-ratings/ (http://projects.fivethirtyeight.com/pollster-ratings/)). Nice thing is, he computes a mean bias of each poll (like a D0.6 means that if the "D" actually has 40%, it would typically show 40.6% in the poll). I used these numbers to compute an adjusted spread. Overall, the adjusted spread average is Clinton +6.4 (+0.2 above the raw average).
I also took a swing at creating a composite MOE for these polls. This is where you guys can back up my stats education...I took each MoE an halved it. Most polls say they are accurate within the MoE 95% of the time, which means they should be reporting +/- 2 Standard Deviations. So, and MoE of 4 gives me a standard deviation of 2. So, I computed the square root of the average of the squares of the Standard Deviations (remember, stats class is a bit fuzzy for me, but this is what I recall the process being), and came out with an composite SD of 1.68, giving a composite MoE of +/-3.4
Overall, this puts the spread at Clinton +3.0 to Clinton +9.7 right now.
Again, just crunching numbers here guys. Looking forward to corrections!
-
You're going to hate me for this, but there is only one poll that matters. Its held on the first Tuesday in November.
-
You're going to hate me for this, but there is only one poll that matters. Its held on the first Tuesday in November.
Yeah, I concur. But it's interesting to compare and contrast stuff.
-
You're going to hate me for this, but there is only one poll that matters. Its held on the first Tuesday in November.
or, to be really nitpicky, it's the one in December.
-
The latest Reuters poll claims the current race has Hillary Clinton at 47 percent support and Donald Trump with 33 percent support.
Reuters interviews 1201 respondents.
626 Democrats (52% of total)
423 Republicans (35% of total)
122 independents
30 other party.
That’s nearly 33 percent more Democrats than Republicans.
In reality Gallup reported in March that 46 percent of Americans are Democrats, and 40 percent are Republicans. Reuters freighted their poll with 20 percent more Democrats than Republicans.
Since the Reuters poll sampled more Democrats than all the others combined, we can safely say that Trump appears to be in much better shape than the poll suggests and could likely be headed to a landslide victory in November.
http://www.eutimes.net/2016/06/reuters-caught-faking-polls-again-hillary-has-no-lead-in-reality-hillary-trump-are-tied/
-
Interesting...64 views of this thread and guess how many people have looked at the actual data...1
-
Interesting...64 views of this thread and guess how many people have looked at the actual data...1
Statistics is not my forte but I'll take a look when I can.
-
Interesting...64 views of this thread and guess how many people have looked at the actual data...1
Why would I look at data if it I don't believe it is going to support my decision? Why should I be different than anyone else?
-
The latest Reuters poll claims the current race has Hillary Clinton at 47 percent support and Donald Trump with 33 percent support.
Reuters interviews 1201 respondents.
626 Democrats (52% of total)
423 Republicans (35% of total)
122 independents
30 other party.
That’s nearly 33 percent more Democrats than Republicans.
In reality Gallup reported in March that 46 percent of Americans are Democrats, and 40 percent are Republicans. Reuters freighted their poll with 20 percent more Democrats than Republicans.
Since the Reuters poll sampled more Democrats than all the others combined, we can safely say that Trump appears to be in much better shape than the poll suggests and could likely be headed to a landslide victory in November.
http://www.eutimes.net/2016/06/reuters-caught-faking-polls-again-hillary-has-no-lead-in-reality-hillary-trump-are-tied/
That's why I like the Real Clear Politics average, which weighs recent polls.
-
That's why I like the Real Clear Politics average, which weighs recent polls.
According to my data, Reuters uses Ipsos for their polling, which as a mean bias of D+0.1 over previous actual election cycles.
-
Any presidential poll which does not do a state by state breakdown of the electoral vote is pretty much useless. It doesn't matter how much Hillary wins New York and New Jersey by, she doesn't get any more electoral votes.
Right now the polling isn't there for a state by state comparison. The "best guess" by the left leaning organizations tend to favor Hillary by ~150 votes, but in some cases they're basing that on "well last time that state was blue".
-
Latest poll from Rasmussen Reports shows Trump in the lead:
http://www.rasmussenreports.com/public_content/politics/elections/election_2016/white_house_watch
I didn't look too far into the numbers. The bigger point is that the poll numbers are going to go up and down for both candidates as we head into the general election season.
-
Poll internals are fun too. For example, in the latest McClatchy-Marist Poll of 1,132 adults, Donnie is running in 4th place among 18-29 year olds, well behind both Johnson and Stein (https://www.scribd.com/document/320225575/McClatchy-Marist-Poll-National-Nature-of-the-Sample-and-Tables-August-2016)
-
http://www.electoral-vote.com/
I think this is one of the better websites for state by state polling. The web master and editor has a very distinct left lean and sometimes I think that colors their willingness to accept "unfavorable" polls. Based on current polling they are showing Hillary winning by about 90 electoral votes. I think their polling tends to be lagging too. In critical states, they can be weeks behind if no polls were done. For example, the last Colorado poll they have is from early July.
http://projects.fivethirtyeight.com/pollster-ratings/
This is an interesting site because the it actually directly addresses the original question - how good are the polls. It turns out, some of them are really, really good at making predictions. Some have left leans, some have right leans. The site tells you which way they tend to go and how good they are at accuracy. So for example, if you hear from Research2000 that Hillary is winning Pennsylvania by 2 points, you can look up and find out that Research2000 has about a 4 point left lean and gets an F on accuracy. From that you can posit that the actual number is likely to be that Hillary is losing by 2 points.
And no - polls are not the end all, be all. But the science behind polling tends to reflect reality IF the polling organization has worked aggressively to remove human induced error in the polling methods. Many years ago as a math major in college I took a statistics course which was all about proving the scientific validity of statistics - which is really all that a poll is, a statistical sample. I could not repeat any of the proofs today but I remember enough to know that there is hard science behind polling and polls are not just guesses.
-
Jesus...
(http://www.pilotspin.com/index.php?action=dlattach;topic=1043.0;attach=235)
-
Jesus...
(http://www.pilotspin.com/index.php?action=dlattach;topic=1043.0;attach=235)
Well done, you angry voters. Well done. You sure showed the establishment that you weren't going to take it any more.
-
Well done, you angry voters. Well done. You sure showed the establishment that you weren't going to take it any more.
To be fair, polls have been going back and forth since the convention (before that, actually) and will continue to do so. I'm not sure that this reflects anything more than one poll from one point in time (today). There are other polls out there showing Trump ahead of Clinton.
-
To be fair, polls have been going back and forth since the convention (before that, actually) and will continue to do so. I'm not sure that this reflects anything more than one poll from one point in time (today). There are other polls out there showing Trump ahead of Clinton.
Actually, it represents a number of polls, all weighted for quality, accuracy, and age.
Can you cite one poll since the end of the Dem convention that shows Trump ahead?
-
To be fair, polls have been going back and forth since the convention (before that, actually) and will continue to do so. I'm not sure that this reflects anything more than one poll from one point in time (today). There are other polls out there showing Trump ahead of Clinton.
At this point, any popular poll means shit. The only polls that will really mean anything, short of the actual vote, are state polls that give an indication as to where that state's electoral votes are going. I have yet to see a poll where Trump got higher than 191 electoral votes, and most show him at much less than that, consistent with the one above.
-
At this point, any popular poll means shit. The only polls that will really mean anything, short of the actual vote, are state polls that give an indication as to where that state's electoral votes are going. I have yet to see a poll where Trump got higher than 191 electoral votes, and most show him at much less than that, consistent with the one above.
Most polls also showed him not earning enough in the primaries to get the GOP nomination, or at least prevent a contested convention. I put little stock in anyone prophesying at this point.
-
At this point, any popular poll means shit. The only polls that will really mean anything, short of the actual vote, are state polls that give an indication as to where that state's electoral votes are going. I have yet to see a poll where Trump got higher than 191 electoral votes, and most show him at much less than that, consistent with the one above.
My point was that polls in general this far out don't necessarily mean anything. I agree that state polls are probably more accurate, but there is still a long campaign ahead. Anything can happen and change the polls. Of course I expect Trump to keep saying dumb things and hurting himself further, but that doesn't mean Clinton won't do something stupid, either. Then there's always the October surprise, whatever that will be this time around.
-
My point was that polls in general this far out don't necessarily mean anything. I agree that state polls are probably more accurate, but there is still a long campaign ahead. Anything can happen and change the polls. Of course I expect Trump to keep saying dumb things and hurting himself further, but that doesn't mean Clinton won't do something stupid, either. Then there's always the October surprise, whatever that will be this time around.
It will be Hillary's health. Count on it. They are already setting it up.
Do you the think they would actually allow a photographer to photograph it taking two men to help her up a short flight of stairs if it wasn't planned?
-
Can you cite one poll since the end of the Dem convention that shows Trump ahead?
I'll take your silence on this as a "No, there haven't been any."
-
It will be Hillary's health. Count on it. They are already setting it up.
No, I think it will be Trump's health.
He'll realize that he has zero hope, and instead of becoming a "loser" he'll suddenly develop a serious, but definitely treatable, condition that will require him to withdraw. And, he'll of course claim that he would have won had he stayed in the race.
-
I'll take your silence on this as a "No, there haven't been any."
I was looking for the one I saw recently, but it appears that it was during the convention. My point still remains; polls are fluid and can change at any point this far out.
-
This close to the election (about 80 days), the polls have about a 75-80% accuracy rate and they will slowly improve over the rest of the election.
Not sure where the electoral map above came from but is appears heavily slanted. I haven't seen one like that for months.
-
This close to the election (about 80 days), the polls have about a 75-80% accuracy rate and they will slowly improve over the rest of the election.
Not sure where the electoral map above came from but is appears heavily slanted. I haven't seen one like that for months.
Since I have Jeff on ignore I don't look at his garbage he post, but if it's from him of course it's heavily slanted.
Major "one trick pony" boy for sure.
-
Yes, well any electoral map that says The Crook will beat The Clown in VA, NC ,SC and GA is a Leftist pipe dream.
-
No, I think it will be Trump's health.
He'll realize that he has zero hope, and instead of becoming a "loser" he'll suddenly develop a serious, but definitely treatable, condition that will require him to withdraw. And, he'll of course claim that he would have won had he stayed in the race.
Maybe... Since the Donald had already declared the election rigged, another scenario is for him to just declare that he won't take part in such a tainted process. Nearly everyone has been unfair, biased or otherwise disagreeable to his conceptual ideas, although it is difficult to keep track of what his position really is on the major issues.... except it will be great.
-
More wet dreams of Trump dropping out.
-
Hillary is only being kept in this to secure a Democrat in the White House, to continue the Soros/Obama Fundamental Transformation of America.
My prediction is that following a Hillary win, she will soon become "debilitated"and Kaine will be sworn in. You heard it here first.
BTW, if I suddenly die with a pillow over my head or am shot dead in a "robbery" where nothing is taken from me, I was killed by Clinton's goons.
-
Hillary is only being kept in this to secure a Democrat in the White House, to continue the Soros/Obama Fundamental Transformation of America.
My prediction is that following a Hillary win, she will soon become "debilitated"and Kaine will be sworn in. You heard it here first.
BTW, if I suddenly die with a pillow over my head or am shot dead in a "robbery" where nothing is taken from me, I was killed by Clinton's goons.
Crickets from the MSM on that Clinton (inspired?) hit.