2016 Postmortem
Related: About this forumAren't you LOVING this trend-line of polling in New Hampshire?
Here is the polling trend-line in New Hampshire:
Here is the same polling graphed with the smoothing minimized (to show the individual ups-and-downs in the polls rather than the general trend lines):
ViseGrip
(3,133 posts)Godhumor
(6,437 posts)Why not show Pollster's default graph for your selected time period:
Oh wait, I know exactly why you don't want to show that one...
Attorney in Texas
(3,373 posts)with smoothing maximized (to highlight trends) and with smoothing minimized (to highlight the more specific ups and downs in the polling).
Godhumor
(6,437 posts)The default moderate smoothing is to show what Pollster feels is a more accurate representation of the current state of the race and the trending to that point.
Forcing all data to a linear trend is done for one reason only. To pretend a race can't and doesn't shift, even over the course of months.
SidDithers
(44,228 posts)you'll notice a "trend" of presenting data in a way that is misleading.
This thread, for example: http://www.democraticunderground.com/1251887694
Whether the poster conflates national and state polls, and implies something that simply isn't true.
Whether that's intentional or not, who can say?
Sid
Godhumor
(6,437 posts)But I don't know if the more smoothing option will pick up on my change in thinking.
Attorney in Texas
(3,373 posts)Godhumor
(6,437 posts)You ever fix those links, by the way?
Attorney in Texas
(3,373 posts)polling.
If you were able to follow my links in order to dispute my post, you were not misled. You just disagree with my conclusion.
By the way, have you found any polling which calls into doubt the conclusion that Sanders does better in head-to-head match-ups with Rubio than Clinton does?
SidDithers
(44,228 posts)Rubio / Sanders - Rubio +6
Rubio / Clinton - Rubio +2
The one Republican who trails Clinton by a wider margin is Carly Fiorina at 46/41. Beyond that Clinton leads everyone by 1 or 2 points- it's 46/45 over Ben Carson, 45/44 over Donald Trump, 46/44 over Ted Cruz, and 43/41 over Jeb Bush.
Bernie Sanders does on average 6 points worse than Clinton in comparable general election match ups. He trails Fiorina 42/40, Bush 42/39, Trump 44/41, Cruz 44/39, Rubio 44/38, and Carson 46/39.
There. There's a national poll, less than a month old, that shows Clinton better than Sanders in a head-to-head matchup with Rubio.
Sid
Attorney in Texas
(3,373 posts)misleading because it does not mention all polling on the topic by every possible source?
Are you deliberately misleading by neglecting to mention and link the subsequent Quinnipiac poll that shows Sanders beating Rubio or the subsequent Clinton-vs-Rubio poll that shows Rubio crushing Clinton by 8% or the SECOND subsequent CNN poll that shows Rubio beating Clinton or the THIRD and most recent USA Today polls that AGAIN show Rubio beating Clinton.
That seems to be your definition for misleading. It's not mine.
SidDithers
(44,228 posts)Sid
Godhumor
(6,437 posts)Last edited Fri Dec 11, 2015, 01:32 PM - Edit history (1)
Your wanted to compare a real aggregate for Clinton vs a single source aggregate for Sanders.
SidDithers
(44,228 posts)on multiple occasions over the past couple weeks.
Sid
Attorney in Texas
(3,373 posts)Hampshire, my analysis will be shown to be inaccurate.
All my analysis predicts is that (1) Iowa looks pretty close and (2) Sanders looks good in New Hampshire and (3) the results in Iowa and New Hampshire will have an effect on subsequent primaries and caucuses but (4) Clinton is still the favorite and Sanders is still the underdog.
If we have a result that disproves one of those conclusions from my analysis, I guess you will be able to say my analysis was inaccurate. Until the, we'll just have to wait and see.
SidDithers
(44,228 posts)You've misused averaging to present an inaccurate but rosy picture for your candidate, you've conflated national and state polling, to present an inaccurate but rosy picture for your candidate, and now you're mangling trend lines to again present an inaccurate but rosy picture for your candidate.
Why don't you do a trend analysis on those datapoints.
Sid
Godhumor
(6,437 posts)Last edited Sat Dec 12, 2015, 09:50 AM - Edit history (2)
Down thread are screenshots of a regression run using five total points of data.
And people wonder why statistical analysts drink...
DCBob
(24,689 posts)There is a pattern with this person.
MADem
(135,425 posts)Over and over and over again....a blatantly poor and misleading effort at using "figures" and "graphs" to paint a picture that isn't consonant with the facts to hand.
Can't say I'm surprised, though.
Attorney in Texas
(3,373 posts)Godhumor
(6,437 posts)Moderate smoothing is also called trending. Linear trending is a forced fit that assumes races cannot change even over the course of months.
You only use more smoothing when you're so desperate to prop up a candidate that you're willing to ignore that races change.
Attorney in Texas
(3,373 posts)body of the post (plus I also posted a second graph with smoothing minimized).
If you don't find the data relevant, you are not obliged to comment on it.
MohRokTah
(15,429 posts)it is right and proper to comment on misrepresentations of data.
staggerleem
(469 posts)... even the graph that's forced linear, "to pretend a race can't and doesn't shift", somehow shoes that it shifted!
Godhumor
(6,437 posts)What I meant about shifting is that, in reality, change didn't work that way.
staggerleem
(469 posts)... but it DID happen, right?
Dem2
(8,168 posts)More smoothing is not that relevant unless polling is very consistent over a long period of time, less smoothing essentially plays 'connect the dots' with each polls error margin, eliminating the whole point of using a smoothing function.
AlbertCat
(17,505 posts)I see Sanders is STILL ahead in this one too.
MADem
(135,425 posts)Consider the source, and let him have his fun. I notice, too, that the OP chart doesn't even have "December" anywhere on it!
This, too, will pass!
Ed Suspicious
(8,879 posts)SoapBox
(18,791 posts)Go Bernie!
Helen Borg
(3,963 posts)Hepburn
(21,054 posts)Enthusiast
(50,983 posts)Attorney in Texas
(3,373 posts)coronation squad who whine that any data which does not fit their story line is misleading).
The Des Moines Register has a pretty good track record (although it is not perfect - it is extra tough to poll a caucus), and here is a graph of the Des Moines Register's polling results (it is shown two ways: with moderate smoothing and then, to better capture the trend, smoothed -- I include both so the coronation squad can get twice as pissy about it):
MODERATE SMOOTHING
SMOOTHED
Either way, Clinton's trend is downward and Sanders' trend is upward.
Attorney in Texas
(3,373 posts)MODERATE SMOOTHING (Quinnipiac)
MODERATE SMOOTHING (CBS/YouGov)
SMOOTHING TO HIGHLIGHT TRENDS (Quinnipiac and CBS/YouGov)
SMOOTHING TO HIGHLIGHT TRENDS (Quinnipiac and CBS/YouGov and Des Moines Register)
MODERATE SMOOTHING TO HIGHLIGHT HIGHS AND LOWS IN POLLS (Quinnipiac and CBS/YouGov and Des Moines Register)
Ultimately, I prefer the Des Moines Register polling and I'm not a fan of the type of internet-based polling that CBS commissioned, but whether you look to the Des Moines Register or Quinnipiac or CBS/YouGov or a combination of these pollsters, Iowa looks like a close race. The coronation squad does not want to hear this, but the higher they raise expectations, the more momentum Sanders will take out of Iowa.
Enthusiast
(50,983 posts)Attorney in Texas
(3,373 posts)Pollster Ann Selzer, who conducts the Des Moines Register/Bloomberg Politics Iowa Poll, noted one similarity between the contest today and in 2007. In Selzer's October poll, Clinton was leading the field, as she was then. But Selzer also noted that Obama held a double-digit lead with independents eight years ago and Sanders holds an even larger lead with that group this time.
"On paper you would say Sanders is in a better place" than Obama, the eventual Iowa winner, in 2007, Selzer said. But the question, she said, is whether he has the organization to turn that potential into caucus votes.
Hepburn
(21,054 posts)I will give up chocolate for a year of Bernie takes Iowa!
Faux pas
(14,682 posts)firebrand80
(2,760 posts)I'm confused.
Attorney in Texas
(3,373 posts)A poll is a snapshot in time, not a substitute for the will of the voters.
firebrand80
(2,760 posts)trashing all the "bad news" polls and exhaling the "good news" polls.
For some reason, I'm skeptical that this will happen.
Bubzer
(4,211 posts)Buzz Clik
(38,437 posts)It's polling data, so we understand it's likely flawed and absolutely limited. But you are simply tweaking until you get the desired result.
Here's a tweak for the HRC fans:
How do you like those trends?
Helen Borg
(3,963 posts)Buzz Clik
(38,437 posts)Attorney in Texas
(3,373 posts)One graph with smoothing shows the trend line and one graph with smoothing minimized shows the individual ups and downs of the polls, and this fact is discussed in the OP.
I'm not sure what contortions you would have to do to show Clinton ahead in New Hampshire, but if those contortions make sense and you believe they create a more accurate graph (say, for example, if you eliminated robo-call polls which are notoriously less accurate), I certainly have no gripe with that.
So, what contortion did you have to perform to show Clinton ahead in New Hampshire and why does that contortion make the graph a better illustration of the data?
Duval
(4,280 posts)ConservativeDemocrat
(2,720 posts)...he's a complete laughingstock.
So, look, it's not bad for the guy to be thrown a bone.
- C.D. Proud Member of the Reality Based Community
Attorney in Texas
(3,373 posts)Senator, and New York is just 50 miles from New Hampshire so the voters in New Hampshire are well familiar with Clinton and she's not losing due to a lack of name identification.
Robbins
(5,066 posts)Clinton won NH in 2008 yet now bernie is leading her In NH.I would say they probally know her better but they like bernie better.
MADem
(135,425 posts)You do realize that those two states share much in common? It's not unusual for people from one state to shop and dine in the other on a routine basis.
To try and pretend there's no association between the two is ... well, silly.
Attorney in Texas
(3,373 posts)recall how Clinton campaigned in New Hampshire in 2007-2008 and she won New Hampshire in 2008 back when she had strong net favorability numbers (and not the net negative unfavorability and untrustworthy numbers she has now)?
Scootaloo
(25,699 posts)As you can see, New Hampshire is nowhere near New York!
MADem
(135,425 posts)NH is an homogenous state, much like VT. They're demographically the same. White and full of hunters who love their guns.
NH and NY are not contiguous, either. NH and VT are.
But way to miss the point! Pat yourself on the back.
brooklynite
(94,607 posts)Ed Suspicious
(8,879 posts)so driven to work for Bernie's campaign. Thanks for the motivation.
RoccoR5955
(12,471 posts)Polls can be manipulated to get whatever results one wants. Be it Clinton or Sanders, or O'Malley. Each one could be a winner in a "scientific" poll. The only science involved is the method and frequency of the manipulation..
There is only ONE poll that matters, and that is the one on Election Day, or in this case, Primary Day.
in_cog_ni_to
(41,600 posts)I like it! A LOT!
PEACE
LOVE
BERNIE
Omaha Steve
(99,664 posts)Former First Lady, US Senator, SoS with a billion $ goal can't get above 50% of late. She has trouble right here in River City.
K&R!
OS
Aerows
(39,961 posts)and it stands for Poll!
Samantha
(9,314 posts)I started looking at local Iowa and New Hampshire papers a couple of weeks ago to get the flavor of the primaries right from the home source. It was very therapeutic. One thing that stayed with me is something some of the our DU'ers from Iowa have reported here. It is almost impossible to correctly poll Iowa because so many voters make up their minds at the last minute.
I personally believe Bernie Sanders has a great possibility of winning both.
Thank you for sharing your information.
Sam
DCBob
(24,689 posts)Misuse of a poll chart under the influence of political bias.