2016 Postmortem
Related: About this forumInteresting New Florida Poll Out Tomorrow Morning
NyTimes bought voter rolls from a vendor and are using the names from the lists as candidates for polling.
This is how the pro's do their internal polling.
Downside is, it was performed over last week so there could be a Sick Hillary effect.
Iliyah
(25,111 posts)Loki Liesmith
(4,602 posts)Most pollsters use either random digit dialing or just sampling from a phone book and dialing people. They then ask someone if they are a registered voter and if they are (assuming they answer honestly) use those individuals in some kind of likely voter model.
With actual registerd voter lists you 1) know they are RVs and 2) know if they have voted lately. You are well on your way to a good likely voter model.
Tomorrow's poll will use voting history and statement of intention to vote to determine likely voters.
First part is good for us, second not clear.
This is an experimental poll. They will use the accuracy of this poll to calibrate future turnout models better.
Because it's experimental, I caution people not to freak out if our numbers aren't good.
That's all I know.
Loki Liesmith
(4,602 posts)Matthew28
(1,796 posts)Loki Liesmith
(4,602 posts)Your as likely to see a Trump lead as you are a Hillary lead.
I'll be happy if she's close.
Sunny05
(865 posts)molova
(543 posts)Explain.
Loki Liesmith
(4,602 posts)If it's like many internal polls it could have a wide horizon for gathering quality data.
So some of her bad days may be in the mix.
Also NYTimes has probably been testing and retesting different scenarios.
I'm not optimistic, but only because optimism makes more mistakes than pessimism.
Sunny05
(865 posts)Clinton up 1% with four candidates considered, tied with just the two compared. Here's the link to the NYT / Siena College poll:
http://www.nytimes.com/interactive/2016/09/19/upshot/florida-poll-clinton-trump.html?_r=0
They explain a fair amount of their methodology, far more than most polling sources that make the "news".
Also, further down in the article, they share what the results would have been if NYT/Siena College had made different choices about how to model. Most results (whether from NYT/Siena College actual model or from these other would-have-been model scenarios) had actual or statistical ties. One would-have-been model would have given Clinton as much as a 6% lead. (I think one scenario had a 4% lead for Clinton, but I am not sure about its margin of error.)
By the way, I found comparisons to other very recent Florida polls (within last couple of weeks), and these results are not hugely different. Some of these other Florida polls were taken before (or mostly before) the pneumonia news, while others were taken over a time frame that included the pneumonia. (Dang it, lost the url about FL polls ... check out reputable poll overviews, poll trackers, poll averaging sites.)
Overall, given the extra "special attention" to dt, the ridiculously negative spin on HRC during most/all days of the NYT/Siena College poll, and the fact that this is Florida we're talking about, I think these findings are pretty darned good. But of course I'm keeping my fingers crossed that things will improve even more for HRC in this state!
Thanks again to Loki Liesmith for alerting us to this.
Loki Liesmith
(4,602 posts)is that polling aggregates appear to be a decent proxy for internal polls.
Sunny05
(865 posts)Hoo boy, polls can be a useful tool, if designed and implemented properly. But of course that is so difficult even for the best trained, most objective pollster in something like political/election polling. I feel tired just trying to think through ... registered vs. likely and likely vs. real voting history and demographics ... and ... and ... political/election polling is tricky. And then there's the whole phone thing and internet based polling.
On some happier notes, other crucial states (Pennsylvania and Colorado) looked good last I checked polls (through, say fivethirtyeight). And the betting exchanges (which I just look at for same purposes I look at poll results, and we can't link to them from here, anyway, since betting on election illegal in U.S.) still favor HRC, though its been closer in recent weeks.
Well, I've said enough. Have a great day.
CajunBlazer
(5,648 posts)Nate Silver has corrected this poll for bias and is using the result of Trump leading by one and he is giving it a very high weight when combining it with other Florida polls.
http://projects.fivethirtyeight.com/2016-election-forecast/florida/
Sunny05
(865 posts)Thanks for the feedback about fivethirtyeight. I am about to check it out through your link there.
A note ... I hope you and others realize that my positive comment on the results is couched in terms of that comment's entire sentence (please see my post above): MSM largely gave dt a good week, MSM and some generally in public (it seems to me) gave HRC a hard time, and my gawd it's Florida. So, of course I'd like for her to have a lead in FL now -- and a huge lead everywhere all the time. But couched in the terms of my post, I think there must be a lot she has done right in order for the poll to be even tied (plus or minus one or a few) given the conditions. And to extend the thought, it seems possible she can and will do even better in the polls without the kind of week she just had (or, perhaps, had flung at her).
Indeed, though, the reality of the election is something I'd describe in a different way, not in terms of "my gawd, given what kind of week was thrown at her, she still held on to this much in FL".
Thanks a bunch. Take care --
Loki Liesmith
(4,602 posts)Up one in the poll, add it to the average.
Nate's bias corrections are conjecture based on relationship to other current polling.
Some people just want to have their little freakout. We have to let them.
CajunBlazer
(5,648 posts).....but Silver's corrections for bias are not "conjecture" - nothing he does is based on conjecture. The points added or subtracted from the poll leaders for bias are based on statistical calculations of not only the relations to other polls, but most of all the bias shown present in a polling company's samples as illustrated by their accuracy in calling past elections.
Sunny05
(865 posts)and perspective. I think we're all frustrated over craziness and unfairness of this election season. But I, like you, liked the result -- given the circumstances I described.
Thanks --
CajunBlazer
(5,648 posts)Just trying to emphasis Florida is not in the bag by any stretch. GOTV!!!
Sunny05
(865 posts)... your thoughts and appreciate the concern over election and desire for HRC to win. Indeed, I added clarification as I did b/c I agree election is not in bag & would love for her to have sound leads across the board!
Also, and this is just a technicality (when it comes to comparing polls, not as pertains to election): tie in poll(s), up one or two, down one or two -- all ties or statistical ties.
Bigger point to me, given context previously described, I was relieved over those particular FL results.
Biggest points to all of us here: we care. We understand what this nation needs, and we care about what it needs.
CajunBlazer
(5,648 posts)For one poll a single point either way is not really significant - however, when you start combining polls based on weighting after correcting for known biases, you have essentially a much larger sample size and less probability of error. Errors in some polls are usually evened out by the opposing errors in other. That is what Nate Silver is doing for every state including Florida.
Right now Silver's combined polls show Trump winning by 0.4% (46.4% to 46.0%) and he gives Trump a 52.8% probability of winning to Hillary's 47.1%. That is at best a toss up now compared to a the 6.3% lead Hillary once held in Florida with a 79.9% chance of winning.
Sunny05
(865 posts)I hear you. I was responding to a thread that started with a reference to that one FL poll. That right there -- combined with the week Hillary had had (thrown at her) plus much unbalanced media coverage for quite some time now -- this specific thread on this one specific poll of this one state in the context of recent campaign/media history is all my post was about.
It sounds as if we want the same thing out of this election. We're on the same side. Yesterday, I responded in manner that acknowledged the concern, etc., of the bigger scheme of things, and attempted to clarify miscommunication and tried to do so in a manner that "connects" people. Again, we want the same thing, I think, as far as the election goes. But please look back at the thread this is a part of. If you want a bigger discussion of all the polls, etc., this is the wrong thread, and this particular response I posted is the wrong response for that. I'm talking one apple and you're arguing an entire orchard. And again, for God's sake, we're on the same damn side.
CajunBlazer
(5,648 posts)But understand that when you say that and apple from an orchard is delicious and without blemish, you are implying without saying so that it is an example of the apples through out the orchard. Hence when you talk about the apple, you might expect that someone will want to discuss the orchard as part of the conversation.
However, you are correct: We are on the same damn side!
Sunny05
(865 posts)I am not implying anything about the other apples. No. Nothing. This thread was an fyi about one specific poll about to come out. After that poll came out, I saw the results of that one poll, then checked this thread to see if anyone had posted its results and/or link, and then posted a quick summary & link. That's it. That is and always was it.
At another time and in another thread, I may discuss multiple polls. But I don't want my particular posts here on the topic of this thread applied to anything else.
Anyway, I DO APPRECIATE very much the fist-bumping smiley faces. And yes, we are on the same side -- which is the sane side.
Take care.
CajunBlazer
(5,648 posts)Reference: "The Blind Side"
Loki Liesmith
(4,602 posts)My God.
Learn to deal.
CajunBlazer
(5,648 posts)the capitalization in the title was not because I am overly concerned, it was meant to emphasis my message and it looks like it worked.
Loki Liesmith
(4,602 posts)No social filter.
The panic attacks I see around here are starting to make me a little nuts. Apologies for unloading on you.