Yougov poll miraculously polls 103% of participants -- Aug 28-Sept 1

[magical thyme]

Here is the email that I've just sent to the polling expert at yougov who produced these stunning results!

Dear Ms. Frankovic,

I'm very curious about the methodology used for the above referenced poll.

Specifically, how did you end up with a total of 103% participants voting? The last I understood, per cent means, roughly, per 100.

For your reference, I am looking at the table that appears in the article at this link:
https://today.yougov.com/news/2015/09/17/clinton-top-second-choice-both-biden-and-sanders-s/

Thank you and regards,

MT

Funny how that table shows support for Bernie dropping precipitously in just 2 weeks. Maybe because the earlier poll added 10 points to his national numbers with zero people undecided and some number of extra people somehow included? Inquiring minds want to know...

[Godhumor]

1. It was a typo

The press release doesn't mention a drop (it says the race is unchanged). The person who made the graph entered 35 instead of 25, which was his level of support on 9/1.

[magical thyme]

4. and some DUers were claiming a sharp drop in Bernie's numbers. Just correcting the record in a way

that hopefully won't get lost.

[upaloopa]

2. A DUer pointed out to me that it isn't a

scientific poll. I am an old statistics student and tutor. I think you need to understand that if the poll isn't made up of a random sample, meaning all persons in a population have the same chance of being selected, then the findings of the sample do not represent the opinion of the population.
This poll doesn't seem to be made up of a random sample from what I have been told.
I still support Hillary but I want to know where we truly stand.

[magical thyme]

3. as a statistics student and tutor, I'm sure you understand that regardless of the randomness or

number of people polled, the *percentage* cannot exceed 100. Right?

And yes, yougov is not a random sampling or a very useful poll.

Personally, I'm less interested in national numbers than state numbers, since results by state are what matter.

(ps. I passed pre-med level statistics with close to a 100% average. I've forgotten a lot in the intervening years, but the fundamentals still stick.)

[leftofcool]

7. Hello typo!

[magical thyme]

8. tell it to riversedge and some of the posters in that thread...

[TexasTowelie]

9. However, you need to keep in mind

that if there are 6 candidates it is possible that with rounding errors to whole numbers it is possible that could account for the totals adding to 103%.

Even with only two candidates there is the possibility that the totals could add to 101%. For people who are acquainted with statistics the practice is that if both options in a sample have exactly half of one percent then they are supposed to be rounded to the nearest even number so the totals add to 100% (one gets rounded up while the other gets rounded down), but what is commonly done is to round both options up. Example: if one option is at 37.5% and the other option is at 62.5%, then it should be rounded to 38% and 62%; however, it will most likely be shown as 38% and 63% instead.

[magical thyme]

15. 1. extremely unlikely and 2. why are there no undecideds? there are always undecideds and they did

include an "undecided" and "other" options in the poll.

And yes, I'm fucking aware that there is an off chance that all 6 candidates could have come in at x.5 and I fucking know how to round. I noted above that I passed pre-med level statistics with nearly a 100% average. I'm not a fucking moron. I can do simple math.

[TexasTowelie]

16. Yes, I saw that and I did not mean to imply anything negative.

I'm merely offering the possibility of what may have occurred with the results. My apologies since I do consider you to be both intelligent and informed.

[magical thyme]

17. I'm going with occam's razor. most likely a typo.

the simplest explanation is the most likely one.

apology accepted. the opening phrase (what you need to keep in mind) made it seem directed specifically at me. Sorry, to... I'm feeling extra grumpy tonight (been sick for 10 days and my spoiled brat dog disappeared and hid from me for nearly an hour while I ran around in the dark with a flashlight looking for him.)

[TexasTowelie]

18. I hope that you start feeling better soon.

I know what it is like to be ill several days so you have my sympathy. Take care of yourself!

[DemocratSinceBirth]

14. I have had this discussion before

It's not a random sample in the literal sense of the word because everybody doesn't have the same theoretical chance of being sampled. They do try to come up with a sample that is a representative of the population.

[Keefer]

5. Rounding error?

Happens more than you'd think.

[magical thyme]

6. more likely a typo. there would have been some percentage of undecideds.

note that the results on the right total 93%, leaving 7% undecided/unaccounted for.

[Godhumor]

10. The graph only shows candidates. Don't know was 6% Other was 1%

For 100%.

[magical thyme]

13. that is the latest poll result. on the left, there are 103% respondants for the candidates...

[Godhumor]

20. And as we already discussed, it is a typo

Yougov was sloppy, but it is just sloppiness.

[magical thyme]

21. it was the graphic designer. see below.

[FSogol]

11. For everyone that has questions about statistics or polling, just use this handy guide:

The FSogol Polling Legitimacy Test:

1. Does the Poll Results favor my candidate? Yes or No?

If Yes, the poll and its methodology are legitimate.

If No, the poll and polling organization cannot be trusted.

2. Does the Poll Results favor my candidate at greater amounts than other polls? Yes or No?

If Yes, the poll is the most accurate, most important, and the most historically correct even if I have never heard of it before and even if it has obvious flaws or is an internet poll where you can vote hundreds, nay thousands, of page clicks, I mean votes.

If No, return to question one.

[jfern]

12. Sanders had 25%, not 35%

But yeah, some people like to claim Sanders dropped 9 when he really gained 1.

[magical thyme]

19. I heard back from Ms Francovic a short time ago.

"Thanks for pointing out the error! I am passing on your note to the person who produces the graphics. The past survey figure for Sanders is incorrect.
Best,
Kathy Frankovic

Sent from my iPad"


It's the graphics person's fault

I love it. In my previous life, several graphics designers drove me effing insane. I was just thinking about the worst of the lot the other day. Don't know what brought her to mind...