Talk:Opinion polling for the 2017 United Kingdom general election/Archive 3

Archive 1

Archive 2

Archive 3

Archive 4

Yougov MODEL not POLL

'< !--DO NOT REMOVE WITHOUT FIRST GAINING CONSENSUS ON TALK PAGE-- >'

The fact that you feel the need to put this, suggests that you know there's a problem, that these models are at best "controversial" if not down

right NPOV

|style="background:#0087DC; color:white;"|11%

|-

| 29 May–4 Jun

| YouGov

| 53,609

| style="background:#B3E0FF" |42%

| 38%

| 3%

| 9%

| 4%

| 2%

| 2%

Just get rid of this crap, it's not on, and I think you know it.

MacDaddy

Results from the YouGov voting estimates (2 rows with around 50,000 as samples) are not considered as polls in the same context as other polls on this page. They need to be removed or placed elsewhere. On top of that, YouGov are not claiming the polls had 50000 people, just the model that includes interviews. 90.215.121.192 (talk) 15:29, 31 May 2017 (UTC)

Yes, the FAQ on their website is quite clear that it is not an opinion poll... "Is this a poll?, No – it is a Multi-level Regression and Post-stratification (MRP) model. Polls ask a representative sample of people a question and then show the results to that question, broken down by demographics." https://yougov.co.uk/news/2017/05/31/yougov-election-model-q/

Read on. The model is based on data from a poll; I don't see what's inappropriate about including the specific poll numbers. Mélencron 16:14, 31 May 2017 (UTC)

The voting intention figures are not the output of an opinion poll, which is the case for all the other polls listed on this page, it is the projection from their model. — Preceding unsigned comment added by 2607:F140:400:A02D:7D75:1101:D43:ED6D (talk) 16:16, 31 May 2017 (UTC)

Also, it's a DAILY MODEL based on an OLD BINARY SAMPLE. The underlying poll yesterday needs to be included but nowhere do Yougov say it was based on 50,000 people. What is so difficult to understand here?? Are you going to be including the Yougov updated MODEL HEADLINE NUMBERS everyday based on ONE OLD SAMPLE? It's absolutely batshtcrazy. 90.215.121.192 (talk) 16:19, 31 May 2017 (UTC)

the two yougov 50k / 53k results should be removed from this table. they are not proper polls. AllCre (talk) 16:41, 31 May 2017 (UTC)

They are proper polls. The question is whether their data sets overlap with any of the other YouGov polls, and there I'm not clear. Bondegezou (talk) 17:00, 31 May 2017 (UTC)

Let me recap:

# Yougov conducted a normal poll which gave results of 42/38 CON/LAB. This should be included as per other polls. The number of respondents for this usual poll they do was NOT 50,000. This is the number of the interviews conducted to build their proprietary model for the seat projection.

# The usual poll they do TOGETHER with the 50,000 interviews were then used to build their seat projection model which is NOT a poll as clearly stated by Yougov themselves.

# This model will then be updated daily up to the election day using the OLD POLL SAMPLE, the 50,000 interviews and any new samples. This daily change in effect becomes a poll of polls seat projection of sorts.

Therefore, the initial underlying normal poll must be included but the number for it is not 50,000. Nonetheless, the daily change in percentages in the seat projection model must NOT as it is not a poll in the same sense as the others in the table. I see that today's update has been removed which is good. The only thing left is to find out the ACTUAL respondent number for the 42/38 poll. I really hope that makes sense and people see sense. 90.215.121.192 (talk) 17:11, 31 May 2017 (UTC)

They are not proper polls, the FAQ on their website even states that they aren't polls. — Preceding unsigned comment added by 2607:F140:400:A02D:7D75:1101:D43:ED6D (talk) 17:41, 31 May 2017 (UTC)

I agree, they are not proper polls, and are distorting the graph, in a way that looks like they are being presented here to influence perceptions of support that will be reported elsewhere. Moreover, the potential damage that the inclusion of such voodoo polls - damage to both YouGov's already lefty reputation, and damage to the polling industry in general is not helpful to those of us interested in actual data rather than politicisation of data via manipulation of data to push a particular political agenda. It is in the interests of all, following recent years' polling catastrophes, that such blatant attempts to distort polling aggregates are challanged head on and corrected; it should be obvious to anyone with a real grasp of data that diverging trends is an alarm bell for dodgy data - DELETE THE YOUGOV MODELS IMMEDIATELY! MacDaddy

If this crap from yougov is going to be updated.. the table of polls is literally going to just be mostly yougov for the next week. They do not belong in this table. Any yougov poll without providing the proper datatables like normal polls, should be removed. AllCre (talk) 17:45, 31 May 2017 (UTC)

Yes but 1et's keep it sane. No reason to get angry. The updated model headline numbers have been removed for today and I hope they will not be added on future days. The initial poll is fine for inclusion though, we just need to find the specifics such as numbers. There is no requirement per se that datatables be published as long as the poll is published in a reputable source and is shown as a weighted daily/weekend/week poll. 88.98.211.79 (talk) 18:18, 31 May 2017 (UTC)

It's not fine, it's distorting the truth... I call NPOV MacDaddy

This is like other discussions on rolling polls and averages conducted during other elections (see Portugal and Canada 2015 or France 2017). Practice has been to keep them, so I think there's little discussion on this. As long as they're data extracted from polls, they're valid. You can put a note to acknowledge their special characteristics, though. Impru20 (talk) 19:06, 31 May 2017 (UTC)

They are not normal for the UK, and everyone knows that. It's bad enough that Yougov churns out far more polls than the others that already distorts any aggregate of polls; but this new level of distortion is too blatant, and going way too far. If you want that data, put it in a separate table and graph for "rolling polls", keep it away from the normal polling. MacDaddy

Then why are we not using the FT Poll of Polls?? Because they are not representative of the others and we have our own moving average. Also, this is a SEAT PROJECTION MODEL using an OLD SAMPLE (the 42/38) as starting point and tweaking it every day to update the MODEL. It's not a poll or even a poll of polls. What is so hard about people to understand here?? It's like banging your head against a brick wall. 90.215.121.192 (talk) 20:18, 31 May 2017 (UTC)

Because a poll of polls is a sort of average (either weighted or whatever) not directly extracted from poll data. This is not that. This a rolling poll model published by a pollster; similar polls have been added for other elections, so this is a settled practice. FT is not a pollster, and (again) a poll of polls is not the same as what YouGov presents here. That you come here and say that this is not a poll just because you say it brings little. This model is directly extracted from interview data, so while not a "poll" like the others, it's not a poll average like the one FT publishes, but actual voting intention data published by a pollster. And you could (maybe) be more respectful to people when addressing them, I think. Impru20 (talk) 20:27, 31 May 2017 (UTC)

I've added a note to these in order to highlight that these do use different mechanics than those used with other polls. Nonetheless, rolling averages in other countries also use different methodologies than "normal" polls and they're still added, given that these are actual poll results conducted by a pollster, so I can't still find the logic behind depriving readers of such an info. Impru20 (talk) 20:41, 31 May 2017 (UTC)

They are releasing polls independent of their model. These are the ones that should be reported - https://twitter.com/SamCoatesTimes/status/870025022760513536 — Preceding unsigned comment added by 2607:F140:400:A02D:50B6:12C4:7FB5:48BE (talk) 21:14, 31 May 2017 (UTC)

SurveyMonkey poll

This poll has not been conducted in the proper scientific way and should not be included in the listAwoma (talk) 21:35, 31 May 2017 (UTC)

1. Give out a source that says they're not done "in the proper scientific way".

2. UK Polling Report as already covered them as a genuine poll.

3. This issue is already addressed in another section of the talk, so let's stick to that one. Impru20 (talk) 21:43, 31 May 2017 (UTC)

How often is the graph updated?

The graph is a nice feature of this page. Would it be possible to update it more frequently? Crundgezhnebv (talk) 21:36, 31 May 2017 (UTC)

I try to update the graph as soon as new polls come out, which is normally within a day. I have been away for a couple of days, so haven't been able to do this, but should be able to update it every day until the election from now on. Absolutelypuremilk (talk) 10:47, 1 June 2017 (UTC)

Thank you Absolutelypuremilk for your great graph work :) [edit forgot my signature] Robertshippey (talk) 12:20, 1 June 2017 (UTC)

Shorter time-span graph

Does anyone else think it would be a good idea to have an additional graph that starts from when the election was called? This would allow for a better visual representation of the changes in the most recent polling. The current graph shows two years and is good for a long term overview but it gets "cramped". The smaller parties in particular aren't very clear. @Absolutelypuremilk: would this be easy to do? Gaia Octavia Agrippa ^Talk 20:15, 1 June 2017 (UTC)

See Talk:Opinion polling for the United Kingdom general election, 2017#Shorter period graph. ;) Absolutelypuremilk (talk) 21:57, 1 June 2017 (UTC)

Oh my... I don't know how I missed that! Gaia Octavia Agrippa ^Talk 22:11, 1 June 2017 (UTC)

No problem, it did make me chuckle! I have added the shorter period graph, feel free to discuss size or positioning here. Absolutelypuremilk (talk) 22:27, 1 June 2017 (UTC)

Let me get this straight...

Under weighted polls, be they daily/weekend/weekly, we are now including proprietary projection models whose major basis is an initial poll (which should be included btw) and further interviews which will be conducted daily to tweak the initial poll. This is absolutely nuts so I won't bother anymore trying to make sense here. There is a case to include the tweaked Yougov poll-initiated interview-based seat projection model somewhere else but not in the general poll table. 90.215.121.192 (talk) 21:34, 31 May 2017 (UTC)

I'll let you straight: don't start new sections in order to address issues already under discussion in another section of this very same talk. There's already a section discussing this very same issue, so stick to that one. Impru20 (talk) 21:42, 31 May 2017 (UTC)

Agreed this talk page is becoming a mess, due to folks repeatedly starting the same discussions under new titles. -- BOD -- 22:01, 31 May 2017 (UTC)

The Yougov poll-initiated interview-based seat projection model we are including now on a daily basis. I see today's rolling tweak hasn't been added btw so much for consistency, is predicting Amber Rudd will lose her seat and a 5000 majority to Labour in a seat that has 6000 UKIPers. What an absolute joke of a "model" and what a shambles for Wiki to include its tweaked headline numbers on a daily basis based on an old, non-constituency sample poll. 88.98.211.79 (talk) 17:23, 1 June 2017 (UTC)

No response then? Good stuff. Well done for messing up the poll table. 88.98.211.79 (talk) 01:06, 2 June 2017 (UTC)

YouGov old poll sample-initiated, interview-tweaked daily model, not poll

If this is shown to be utter BS by the actual result after having clogged up the poll table daily with tweaked repetition, YouGov should NOT be used for any future UK elections. Add your support to this below. 88.98.211.79 (talk) 01:08, 2 June 2017 (UTC)

As above, we shouldn't include every instance of a tracking poll that uses a moving average approach as they are not all distinct. However, there is no reason to ban every YouGov poll. Bondegezou (talk) 09:02, 2 June 2017 (UTC)

Legend on graph or below graph

The graph by Lmmnhn has the legend at the top of the graph, I think having the legend in the text underneath works better. Thoughts? Absolutelypuremilk (talk) 10:59, 2 June 2017 (UTC)

The y axis lines are too thick that was the reason I changed at the first place. Lmmnhn (talk) 11:51, 2 June 2017 (UTC)

I've changed them now, is that better or still too thick? Absolutelypuremilk (talk) 11:53, 2 June 2017 (UTC)

Oh and I think a line for each 5% or even 10% would be better. The current 2.5% is too much wouldn't you think? Lmmnhn (talk) 11:54, 2 June 2017 (UTC)

I think 2.5 allows people to match the table to the graph easily which is important. But if others agree with you I'm happy to change it. Absolutelypuremilk (talk) 11:57, 2 June 2017 (UTC)

Transparent data points on graph

Lmmnhn has added a new graph with larger but more transparent data points, do people prefer this style of data points to the one on my graph here? Comments on whether a line for the EU referendum should be included are also welcome. Absolutelypuremilk (talk) 08:31, 2 June 2017 (UTC)

While i mostly prefer the thinner line, both graphs have problems when covering the election period, the SNP's yellow line is particularly hard to see.-- BOD -- 08:58, 2 June 2017 (UTC)

I agree but I think that colour matches what the SNP normally use, could you suggest an alternative colour? Absolutelypuremilk (talk) 09:00, 2 June 2017 (UTC)

I do think its good that the graph follows the colours that political parties normally use. And while I think it might be clearer if SNP line and dots where another colour, I am not sure what (The rainbow of colours is so limiting;) ). I guess that's why the second graph is so useful (doh!). -- BOD -- 09:26, 2 June 2017 (UTC)

Maintaining it up to date I guess is more important.-- BOD -- 14:23, 2 June 2017 (UTC)

Update Ipsos MORI poll source post publication

In line with the sourcing of previous polls, I'd like to suggest that we change the source for the latest Ipsos MORI poll to their website publication found here: https://www.ipsos.com/sites/default/files/2017-06/pm-june-charts.pdf

There is also a more technical publication located here: https://www.ipsos.com/sites/default/files/2017-06/pm-june-2017-topline.pdf

That is of course if I'm not missing anything and indeed this is the source in question Alkanna (talk) 15:01, 2 June 2017 (UTC)

Thanks, I have added this to the table. Absolutelypuremilk (talk) 15:14, 2 June 2017 (UTC)

Demographic breakdowns

I saw a report in the news the other day about how people in a certain age group were changing their voting intentions as the campaign progressed. Would it be worth having a section on polling by demographic group (age, income etc)? Sophie means wisdom (talk) 12:32, 3 June 2017 (UTC)

Sure, the fact that there is a large young/old divide has been covered extensively in secondary sources. Absolutelypuremilk (talk) 13:56, 3 June 2017 (UTC)

Ditto. Agreed. Bondegezou (talk) 14:13, 3 June 2017 (UTC)

Semi-protected edit request on 3 June 2017

1. Change the the latest opinion poll link resource (The one simple named "Survation" which has a 1% Conservative lead) to "http://survation.com/wp-content/uploads/2017/06/Final-MoS-Post-BBC-Event-Poll-020617SWCH-1c0d4h9.pdf" 2. Change the name of the poll named in the first point to "Survation/Mail on Sunday" in line with the rest of the polls Alkanna (talk) 21:12, 3 June 2017 (UTC)

Shorter period graph

Any chance of having a graph showing just 6 months to June 2017 as its getting very cluttered on the existing graph on the right hand side. - Ânes-pur-sàng (talk) 08:24, 19 May 2017 (UTC)

I think the changes over time are fascinating and we'd lose a lot by doing this. --Dweller (talk) Become old fashioned! 09:46, 19 May 2017 (UTC)

I'm not thinking of instead, but also..... - Ânes-pur-sàng (talk) 17:27, 19 May 2017 (UTC)

An additional graph could be a useful thing. Anyone want to mock one up and see how it looks? To avoid POV, we probably ought to start it from a concrete point in time, maybe when May called the election, or when Parliament voted for it. --Dweller (talk) Become old fashioned! 08:57, 23 May 2017 (UTC)

It's a great idea. At the moment, all the action is happening in a postage stamp sized area.Boscaswell talk 07:24, 26 May 2017 (UTC)

Good idea, I will mock one up now starting from 18 April. Absolutelypuremilk (talk) 08:43, 26 May 2017 (UTC)

A second graph for the election period would be excellent. - Galloglass 09:24, 26 May 2017 (UTC)

I'm not very good at creating graphs, but Just looking at data for May, without making averages, I got the following. I find the top one is the most interesting of the pair. - Ânes-pur-sàng (talk) 09:46, 26 May 2017 (UTC)

I think all parties need to be in the same graph really, otherwise there are going to get a lot of objections. - Galloglass 10:34, 26 May 2017 (UTC)

Quite true, unless it is being used just to highlight the closing gap between the top two parties. - Ânes-pur-sàng (talk) 11:01, 26 May 2017 (UTC)

Here is the full graph with shortened axes. Absolutelypuremilk (talk) 10:37, 26 May 2017 (UTC)

Polling for the short campaign of the 2017 UK General Election (from 18 April 2017 onwards)

Much easier to read. Is it still using 10 poll averages ? If so, any chance of a 5 poll average as sharp changes tend to lag behind recent polling ? - Ânes-pur-sàng (talk) 11:01, 26 May 2017 (UTC)

Thanks, here is a 5 poll average version. Absolutelypuremilk (talk) 13:08, 26 May 2017 (UTC)

Polling for the 2017 UK General Election from 18th April onwards.

That looks great, clear and concise, well done. - Ânes-pur-sàng (talk) 13:11, 26 May 2017 (UTC)

Any chance of seeing it in the article ? - Ânes-pur-sàng (talk) 17:10, 30 May 2017 (UTC)

Where do you suggest it goes? Absolutelypuremilk (talk) 18:32, 30 May 2017 (UTC)

How about to the right of the contents list, i.e. just above the main graph. Unless it appears on this page, it is unlikely to be added to the main UK General Election page. - Ânes-pur-sàng (talk) 13:27, 31 May 2017 (UTC)

I'm fine with that if you can figure out how to do it! Absolutelypuremilk (talk) 14:40, 1 June 2017 (UTC)

Now I'm in the right place Could you please do an updated version of the 5 poll average? And maybe with a weekly x-axis. Gaia Octavia Agrippa ^Talk 22:18, 1 June 2017 (UTC)

Hey all, I really liked the addition of the shorter period graph, so I thought I'd try to update it. I used a 5 poll rolling average but used a 'mid-point' date for all the polls (not sure if that was how it was done before). I also used the colours for each political party from this website: political party colours. Finally, I also added a bar showing when the election was called and continued the graph up to the election data. Let me know what you think! I've embedded the image into this page, but it is also here: image ZeroRPM (talk) 18:18, 3 June 2017 (UTC)

Graph of polling information for the 2017 UK general election. The time period is from the calling of the election up to the election date.

I like the having the general election date as the end of the graph, that's a good idea. Could the current graph be changed to reflect this? Its best to keep the same style for both graphs. As such, could the line for the short-time graph be made thinner so as to reflect the longer-time graph. Gaia Octavia Agrippa ^Talk 19:11, 3 June 2017 (UTC)

Hopefully the OP for the main graph will come back and can adjust their graph! I've made the lines thinner for the short-time graph now, I'm not sure how the original graph was made but the default sizes/fonts etc seem a little different to mine - I tried to make it as consistent as possible! ZeroRPM (talk) 19:31, 3 June 2017 (UTC)

I made my graph on numbers (on a Mac) so it might be a little different. There isn't actually enough room to fit in both an "election called" line and a "date of election" line (unless I offset the captions, but then it looks a bit strange). I've changed the colours to those suggested. Absolutelypuremilk (talk) 09:24, 5 June 2017 (UTC)

Ahh that must be it (ref:Mac). Just about the colours...I found a 'better' source, as wikipedia has templates which describes the colours used (however I think only the Green party was any different, although maybe Conservative was too):

Party	Colorcode	R	G	B
Con	#0087DC	0	135	220
Lab	#DC241f	220	36	31
UKIP	#70147A	112	20	122
Lib dem	#FAA61A	250	166	26
SNP	#FFFF00	255	255	0
Green	#6AB023	106	176	35

ZeroRPM (talk) 17:08, 5 June 2017 (UTC)

National Seat Projections section

I'd like to suggest two enhancements:

1. An extra column showing the make-up of the House of Commons at the time of the dissolution of Parliament
2. An extra row showing (where not a hung parliament) the majority that projection indicates, or how many short of a majority the leading party is projected to be, to avoid people having to do mental arithmetic

--Dweller (talk) Become old fashioned! 10:02, 2 June 2017 (UTC)

Agrees with both points-- BOD -- 10:06, 2 June 2017 (UTC)

I like this idea - will add. FriendlyDataNerdV2 (talk) 16:28, 4 June 2017 (UTC)

I'm not sure what order the projections are in. NPOV and ease of editing would normally suggest alphabetical but I wonder if chronological would be best for readers? --Dweller (talk) Become old fashioned! 09:32, 6 June 2017 (UTC)

Not a wiki member put there are some more constituencies polled. So feel free to add them

Reported here http://www.gazettelive.co.uk/news/teesside-news/latest-yougov-general-election-polls-13121450

Stockton South Labour 45% Conservatives 44%

middlesbrough south and east cleveland Conservatives 49% Labour 47%

The other polls are graph based and are hard to read. — Preceding unsigned comment added by 2.25.150.67 (talk) 18:28, 2 June 2017 (UTC)

Those aren't constituency polls, those are modeled results based on interviews across Great Britain. Mélencron 18:46, 2 June 2017 (UTC)

So we admit those aren't POLLS, and yet the same YouGov MODEL tweaked daily is included every day. Good to know. Absolute shambles. 88.98.211.79 (talk) 01:48, 3 June 2017 (UTC)

You still haven't got my point, have you? It models constituency results based on a large national poll. Mélencron 02:01, 3 June 2017 (UTC)

I was the original poster, that is somebody else who responded. Thanks I see what you mean. — Preceding unsigned comment added by 2.25.150.67 (talk) 09:12, 3 June 2017 (UTC)

By the way, feel free to sign up for an account. It takes two minutes, costs nothing and we'd love to have you, especially as you seem intelligent and responsive to logical argument. These are rare qualities on the internet. If you do sign up, drop me a line at my usertalk page and say hello. --Dweller (talk) Become old fashioned! 09:36, 6 June 2017 (UTC)

Survation and rounding

The published numbers seem to be CON 41.5%, LAB 40.4%. That suggests that the rounded number should be 42%, no? (Unless it's actually CON 41.48% or something, that is.) Mélencron 00:19, 6 June 2017 (UTC)

Why don't you change it then? All sources have it as 41.5% which is 42% in any wiki poll page ever made. <personal attack redacted by Dweller> 88.98.211.79 (talk) 02:03, 6 June 2017 (UTC)

I also wondered about this. The exact figure appear to be $367/(914-30)$ (page 10) which is 41.51%. So that's 42% after rounding to the nearest integer. It should be noted that survation themselves report this as 42.[1] Rami R 09:44, 6 June 2017 (UTC)

I'm not sure which survey it is you're talking about, but the suggestion seems sensible on face value. Go ahead and do it. --Dweller (talk) Become old fashioned! 09:56, 6 June 2017 (UTC)

DELETION of YouGov 50,000 MODELs

The following discussion is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

---

These are clearly controversial, they are NPOV and do not enjoy support. It is provocative and contentious for someone to keep posting them and to insert comments like this: "DO NOT REMOVE WITHOUT FIRST GAINING CONSENSUS ON TALK PAGE", when they have not actually sought consensus themselves. For this reason, I am deleting them, and the individual who wants to add them should be the one seeking consensus in advance of posting them there, because there is clear consensus of objections to them being there on this talk page. Argue with me if you want, but you can't ignore the fact that there are multiple requests for these polls to be removed, and no seeking of consensus to include them. Macdaddy 12:20, 6 June 2017 (UTC)

Who are you? You've signed this comment as "Macdaddy", but this comment has been made by an IP editor according to the page's history.

On the issue at hand, these're poll data conducted by a pollster, so they must be included in the page, since I don't think it's up to anyone to select the pollster's info that must be published.

I, however, concede that there's a point in this reasoning. I don't support this data removal, BUT I wouldn't oppose that these are noted in a separate table or so. After all, their methodology is different, and they aren't included in the chart anyway. Impru20 (talk) 12:28, 6 June 2017 (UTC)

Who am I? I have posted my user name, I don't log in. I've been on Wikipedia since 2001, if that helps.

It's not good enough to just have "is published by a pollster", you actually have to have comparable and open methodology to put appropriate datasets together, otherwise you're making too many assumptions, and there's a significant risk of comparing apples with oranges. If pollsters don't open up, they don't pass a reasonable threshold of scientific credibility, simples.

They should be in a different table, because they are different data; and that precedent is already established on the page.

It's bad enough that there are already so many ropey polls being published, the post-mortem on this issue seems inevitable.

look at electoralcalculus.co.uk for a more credible aggregation.

12:48, 6 June 2017 (UTC)Macdaddy

Well, then use your account. Sorry, but it's highly dubious for an IP user to keep attributing comments to an otherwise unused user account (Macdaddy's last activity was in 2011, and its first one dates back from 2004, not 2001. If you are really him/her, log in or, in the event you can't because you forgot the password or have no other mean to access it, create another account. Alternatively, just use your IP account. But it seems rather weird that you don't log it, that you claim to be here "from 2001" (as if that had any impact in the discussion, which it actually doesn't) and then that we're unable to prove your claims because of this weird situation.

I've solved the issue by creating a specific, separate section for these models. This way we can have them in without having to mix them with other polls or remove them altogether. I think this would be a decent compromise with everyone. Impru20 (talk) 12:59, 6 June 2017 (UTC)

It seems to me that the general consensus of the most active group of editors on this article is to include everything, regardless of whether or not it meets any kind of standard. If you have a scroll through this talk page, you can clearly see that. Now, I hate to weigh in simply to point this out, but it seems necessary at this point to seek some kind of neutral arbitration on various matters to do with this article.

iMarc89 (talk) 12:45, 6 June 2017 (UTC)

You can clearly see that nothing is clear at all!!! LOL ...raising the notion of "neutral arbitration" is a laugh... how many of you have STEM degrees and work with data? I'm not trying to be belittling, but it does rather affect the kind of conversation that can be had.

12:50, 6 June 2017 (UTC)Macdaddy

Claiming that you're not trying to be belittling when you are engaged in the process makes you come across even worse. Of course, you understand, I am actually taking your side in this. I don't believe this poll should be included as it isn't actually a poll per se.

My point in the beginning was that there doesn't seem to be any kind of standard set on this page for what should and should not be included and that the general consensus of the more active editors is obviously to include everything. However, your most recent reply puts an exclamation point on my suggestion of neutral arbitration. If you cannot separate your anger from your views, there is no point taking to you.

iMarc89 (talk) 13:06, 6 June 2017 (UTC)

Nonetheless, it's clear that the situation is disputed, and that there is indeed some grounds for it. As a result, I've created a specific new section for these models, so that these are not mixed with "normal" opinion polls but they're not removed either. Impru20 (talk) 12:59, 6 June 2017 (UTC)

I made logical, considered points about this last week many times. Include the headline poll, and put the tweaked daily model in another section. The lefty loonies started making up excuses. Seems editors here are not unbiased at all. And they clearly just ignore IP posts even though Wiki is free to use and edit. Instead, new users who just set up a basic account are treated as gods. 90.215.121.192 (talk) 16:09, 6 June 2017 (UTC)

Please read WP:NPA 'lefty loonies' is uncalled for. Dbrodbeck (talk) 17:00, 6 June 2017 (UTC)

If the shoe fits. Look below at my considered points. I even made a bullet point list. Then a user called "Melencon" (definitely not a French lefty loony amirite?) started shutting down every argument with new rules and regulations even though they themselves admitted the YouGov daily tweaked model, whilst based on an initial poll sample, was from Day 2 onwards in their own words a "Multi-level Regression and Post-stratification (MRP) MODEL BASED ON poll data". Good that sanity has prevailed but for the past week the poll table in "neutral" Wikipedia has been a propaganda tool. Well done to all the lefty loonies involved. Chapeau 90.215.121.192 (talk) 18:44, 6 June 2017 (UTC)

I chose the username because I liked the absurd portmanteau, not because I adhered to their views. Please drop the unfounded accusations and personal attacks. (I've now been accused of being a supporter of the FN, AfD, and Fidesz in the past, and for once I'm being accused of being a lefty loon. Refreshing for a change.) Mélencron 19:19, 6 June 2017 (UTC)

Whatevs as they say. Every time someone made logical, considered, common sense points about the YouGov model, you were the first on the scene to shut them down and unilaterally declare it was a poll in the same vein as all the others, including the running YouGov poll itself, which is frankly ridiculous. It is due to people like you that people give up trying to improve Wikipedia. I could have easily set up sockpuppets and forced changes but I thought people here were balanced. I was wrong. 88.98.211.79 (talk) 20:36, 6 June 2017 (UTC)

So now you are threatening sockpuppets?! Please log in and identify yourself. Anjoe (talk) 21:05, 6 June 2017 (UTC)

Indeed, he could be already be using them as per WP:ILLEGIT (at least these two IP accounts, 88.98.211.79 and 90.215.121.192).

Anyway, user seems to be openly violating WP:AGF, WP:NPA while openly threatens to use WP:SOCK. Impru20 (talk) 22:42, 6 June 2017 (UTC)

Are people here wilfully ignorant or just dumb? Read what I just said again, S L O W L Y, and then delete your comments made in haste. I shall do the same for this one. Jesus wept 90.215.121.192 (talk) 00:28, 7 June 2017 (UTC)

Calm down buddy. Read what I said again and slowly and then delete your comment. 88.98.211.79 (talk) 22:27, 6 June 2017 (UTC)

How can an IP editor who we can't track as a specific user be allowed to make such a controversial edit? This should be reversed until there is some arbitration between registered users or an admin makes at least a temporary judgment on the matter. Anjoe (talk) 17:58, 6 June 2017 (UTC)

The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

Deletion of the SurveyMonkey poll.

First of all, please don't delete a poll without a discussion on the talk page. There is no good reason not to include the SurveyMonkey poll. SurveyMonkey is a known pollster. SurveyMonkey polls are included for example: https://en.wikipedia.org/wiki/Nationwide_opinion_polling_for_the_United_States_presidential_election,_2016 There is no doubt, that this is a scientific poll. The poll results are cited in a newspaper, so we have a source. That the comments about the results in the newspaper may be biased doesn't matter, because that doesn't change that the poll results are facts. Only these facts are included in the table. -Ich bin es einfach (talk) 14:34, 27 May 2017 (UTC)

A newspaper is not a credible "source" - newspapers report polls, often erroneously. Stop talking at people as though you are being reasonable, credible, and erudite, you are not - the onus is on YOU to obtain consensus BEFORE posting your favourite poll... you can't just post whatever you want and then forbid everyone else from removing it, that's not how Wikipedia works. I have just removed something because there is a lot of objection to it, and no attempt to seek consensus before posting it, and yet whoever is posting it is attempting to exert authority over everyone else by forbidding removing it, which is totally inappopriate, and totally unwikipedian. pack it in. 12:25, 6 June 2017 (UTC)Macdaddy

The SurveyMonkey poll doesn't appear to come with any published data or methodology beyond what is in the Sun article. Unless this can be linked to, it shouldn't be included in the article. SurveyMonkey doesn't appear to be a polling company at all, but a company that sells market research technology. Formerip (talk) 14:57, 27 May 2017 (UTC)

The Sun says SurveyMonkey conducted the poll. Crundgezhnebv (talk) 15:40, 27 May 2017 (UTC)

'Surveymonkey' is simply not acceptable. No methodology given, no link to the survey data, no data on who it asked (6000 Sun readers would be clearly biased), not a member of any polling organisation, only source is a clearly biased article in a newspaper. It fails on every account. Jw2036 (talk) 09:38, 28 May 2017 (UTC)

SurveyMonkey are noted for not publishing their data, so are not an acceptable source. For this reason we don't normally include polls such as these. If people wish to include them, then they need to establish a new consensus here as the current one is not to include. - Galloglass 15:49, 27 May 2017 (UTC)

"Are noted"? Who by? "So" they're not acceptable? What's the logic? And who says, other than you? Many SurveyMonkey polls are included in the article for the 2016 US presidential election. There's no consensus that they shouldn't be included in articles here. Crundgezhnebv (talk) 16:19, 27 May 2017 (UTC)

For what it's worth I wouldn't have included it except SurveyMonkey did a 2015 poll that made it into the 2015 polling article for a while. Having just checked the 2015 article I see it has been removed by user Jw2036 at 2pm today (who also removed them from this article), so until today there was a precedent for including them. I have restored their presence in the 2015 and 2017 articles and invited the editor to discuss it here. FriendlyDataNerdV2 (talk) 16:26, 27 May 2017 (UTC)

FriendlyDataNerdV2 when was the 2015 survey reinserted originally back onto the 2015 poll page? regards - Galloglass 18:24, 27 May 2017 (UTC)

Here is an article about the SurveyMonkey poll for the 2015 UK election: https://fivethirtyeight.com/features/surveymonkey-was-the-other-winner-of-the-u-k-election/ -Ich bin es einfach (talk) 17:46, 27 May 2017 (UTC)

Galloglass - it was there right up until today when the editor removed it. Before that, as far as I can see by looking through older versions of the page, it wasn't removed once after being added. I noticed the removal today and put it back on. FriendlyDataNerdV2 (talk) 19:34, 27 May 2017 (UTC)

Odd. I believe it was discussed 2 years ago and removed then. - Galloglass 19:37, 27 May 2017 (UTC)

You are right, it was removed and has crept back in since Jw2036 (talk) 09:43, 28 May 2017 (UTC)

The fact remains that they generally do not release their tables/methodology. At the very least I think there should be a footnote or link to the methodology section making this clear so readers can decide for themselves if they want to trust it. ChiZeroOne (talk) 20:18, 27 May 2017 (UTC)

Sure, you can put a note in the article, there is nothing wrong with that. -Ich bin es einfach (talk) 16:32, 28 May 2017 (UTC)

Galloglass - ah ok, I couldn't find a point where it had been removed. Could just be me.

ChiZeroOne - To be fair they did release the 2015 tables, and I'm happy to remove this one until they release the tables, but I feel it's worth pointing out that's not a standard we apply to other pollsters with a record of previous polling, which SurveyMonkey does have. This isn't a big issue for me so I'm happy to go with the majority opinion FriendlyDataNerdV2 (talk) 21:17, 27 May 2017 (UTC)

FriendlyDataNerdV2 - It may well have been a seperate poll by SurveyMonkey. I do remember it being extensively discussed and found wanting against the BPC polling firms that publish their data sets within a few days. - Galloglass 21:37, 27 May 2017 (UTC)

This is like a Groundhog Day debate. See the 'Quriosly' version of it last year on the EU referendum, or similar going back to 2015/Scots indyref. These are market research companies who don't release their methodology or data and who could well have serious biases in the sample they select. They fail on every account as a reliable source. It's amazing to keep coming back to this at every vote! Jw2036 (talk) 10:00, 28 May 2017 (UTC)

Keep. This article is a list of published opinion poll results, it is not and should not pretend to be anything more than a list. There are no accepted criteria for what constitutes a genuine poll so how can people apply such criteria? Stub Mandrel (talk) 10:01, 30 May 2017 (UTC)

Delete. Don't talk such crap. It 'is' more than a list, and there 'are' accepted criteria for what constitutes a genuine poll, and the onus to prove the validity of content is on the person trying to add something to the page, not the other way round. You build consensus, then you add; you don't impose consensus, and then guard a wikipage like a troll under a bridge - I've seen this behaviour on this site for the last 15 years, pack it in. [ 12:29, 6 June 2017 (UTC)Macdaddy

Is that aimed at me @Macdaddy? I'm not 'guarding like a troll', this is only my second talk comment about a page that I haven't edited. Please stop using intimidating language and take a more mature approach; perhaps you should re-read the basic Wikipedia guidelines starting with 'assume good faith'.Stub Mandrel (talk) 08:34, 7 June 2017 (UTC)

Someone has unilaterally deleted the 6 June SurveyMonkey poll. All the earlier SurveyMonkey Polls appear to have been reinstated, why has today's (6 June) one been removed without discussion?Stub Mandrel (talk) 08:34, 7 June 2017 (UTC)

Survation lead

The 2-3 June Survation poll has Con 41.5, Lab 40.4, which would give a Tory lead of 1.1, rounding to 1. However, if you round the Con/Lab scores to 42/40 as in the table, the lead is 2. What should we put as the lead? I think this issue has come up before, I can't remember what we decided to do last time. Absolutelypuremilk (talk) 15:18, 7 June 2017 (UTC)

If we put 42/40, we should put 2. If you want to put 1, then you would have to have it as 41/40. Otherwise it'd just be inconsistent. Impru20 (talk) 15:25, 7 June 2017 (UTC)

Let's also bring user Anon2424 to the discussion. Let's face it: the full consequences of rounding must be assumed whatever is done. It's just incoherent to pretend to two different roundings applied for the lead than for voting intention figures. The Survation poll has a lead of 1.1 because the original numbers are 41.5/40.4. But if you choose to round it as 42/40 (thus subtracting 0.4 points from Labour and adding 0.5 points to the Conservatives), then you can't round the lead backwards (compared to that rounding) to 1. The lead turns into 2% as a result of the rounding, and because 42-40 = 2 (simple maths). If people wish to have a lead of 1, then it should be the voting intention figures the ones that should be rounded to reflect such a lead, but you can't have a table that says that 42-40 = 1, because that's mathematically inconsistent. You can have either the Conservatives in 42% or the Lead at 1%. Not both. Impru20 (talk) 15:38, 7 June 2017 (UTC)

I appreciate you leaving the difference of 1% however rounding the Conservative figure down would also be incorrect. I understand your logic, however there will naturally be rounding issues causing slight differences or percentages not totalling 100%. All figures in the table must be rounded independently, and if this means having 42/40% with a difference of 1% then this is explainable by rounding. What would be misleading, is to round a headline figure of 1.2% up to 2%. This adds almost an extra percentage point to the figure which is material. Anon2424

Who is doing the rounding, is it us or the pollsters? If we have the report, and it has decimal places, we ought to use them I think. That's what we did in the 2015 Canadian election anyway. Dbrodbeck (talk) 15:49, 7 June 2017 (UTC)

The actual numbers are 41.5, 40.4, 1.1. In a rounded table they should be presented as 42, 40, 1. Rounded tables don't always preserve the additive relationships between numbers, which is why you sometimes see notices like "these percentages may not add up to exactly 100 due to rounding." This is OK as long as it is explained. Arbitrarily rounding 1.1 to 2 or rounding 41.5 to 41 is not OK ever. 62.255.245.226 (talk) 15:59, 7 June 2017 (UTC) Absolutely agreed. Anon2424

Anon2424 Firstly, it's untrue that "figures in the table must be rounded independently", but besides that, the lead figure is NOT given by the pollster: it's us the ones who calculate it. Secondly, the rounding is not done by the pollster, but by ourselves. So, the data we depict here (whatever it being 42/40 or 41/40) is not what the pollster actually says, but what we interpret from it. We can do it under WP:CALC, since it's just a mathematical proceeding, but then you can't have it as mathematically inconsistent by showing a lead that is not accurate with the rounding we actually did. So, no. The lead is calculated under WP:CALC, as it's not given by the pollster in its tables, so you can't have it as mathematically incoherent with the poll's data. If you "absolutely agree" that the numbers are indeed 41.5 and 40.4 (the 1.1 one is not given), then maybe these should be the numbers used. As another but related issue, I don't understand where's the need for rounding all numbers in UK polling articles, since this is not done for other countries (I agree with Dbrodbeck here). Impru20 (talk) 16:06, 7 June 2017 (UTC)

I think the most important part of WP:CALC is that the calculation must be a meaningful reflection of the source data. We all now know the source data is x=41.5, y=40.4, yet we are presenting an x-y column with an answer of "around 2". That answer reflects our own rounding error. It would be better to reflect the source data, which would yield x-y = "around 1". I understand the risk of confusing people, but IMO that is best dealt with by adding a footnote next to the 1 in the lead column. 62.255.245.226 (talk) 16:23, 7 June 2017 (UTC)

Or, if we acknowledge that having Conservatives as 42% is the one issue causing such errors, it can be rounded it down to 41% (given that 41.5% if just halfway between 41% and 42%). The issue is that what you say of having the lead as 1% is not a calculation, but the rounding of a calculation over non-rounded numbers that, then, are presented rounded (this is: you calculate everything before the rounding and THEN round everything, even if the result is mathematically inconsistent). A footnote is absurd in these circumstance given that it's us, not the source, the ones causing the issue. The source is very clear on this: 41.5/40.4. It doesn't give the 1.1 figure (which is calculated by ourselves). The issue comes because someone choose to put Conservatives at 42% (+0.5) despite the fact that Labour got down to 40% (-0.4). That +0.9 difference in favour of the Conservatives as a result of the rounding is what causes the distorting. However, given that the distance of 41.5% is the same from 41% than it is from 42% (that is, 0.5 points), either rounding would be valid in this case. So, we can just have it as 41/40, OR just stick at what the source says and don't make any rounding at all. Else, it'd be both confusing to people and incredibly absurd given the circumstances. No matter how many footnotes you try to add to pretend to solve the issue: people will still get confused with a footnote that pretends to say them that this must (somehow) be shown as 42/40 AND a lead of 1 instead of other, simpler and more logical solutions. Impru20 (talk) 16:33, 7 June 2017 (UTC)

I would be pretty happy with just presenting the data to 1 decimal place though. My best guess is that the decision to round was probably made because that's how the figures are usually presented in UK media, but we needn't slavishly adhere to that. I'm no expert on the intricacies of Wikipedia policy & convention(!) but perhaps it would be OK to use the decimal place when we have it. (Another +1 for Dbrodbeck.) 62.255.245.226 (talk) 16:23, 7 June 2017 (UTC)

YouGov moving sample on graph

For the graph, I am unsure how to cover the YouGov moving sample, as obviously including a daily moving sample will drastically shift the moving average towards whatever YouGov's model predicts. Shall I just include the first poll or are there other suggestions? Absolutelypuremilk (talk) 14:29, 2 June 2017 (UTC)

Somewhat tend to agree. For the chart, I guess you could have only the "normal" polls, while the moving sample could be left just for the table to prevent it from drastically influencing the trendlines. The notes used for the table could be edited to report this situation. Impru20 (talk) 14:42, 2 June 2017 (UTC)

Let me see here. We're including the old-sample-initiated, interview-based MODEL from YouGov and its daily tweaks and considering them as POLLS, which they aren't but we are not including them in the averages graphs. This absolute joke gets better and better. 90.215.121.192 (talk) 23:42, 2 June 2017 (UTC)

This is a rolling average which bases on a completely new sample only every seven days. To not give it overweight you can include the first result in the graph and then the result after seven days. -Ich bin es einfach (talk) 09:34, 3 June 2017 (UTC)

Agreed. Bondegezou (talk) 10:37, 3 June 2017 (UTC)

Yeah, agreed, that is what we did with the Nanos rolling poll in the 2015 Canadian election. Dbrodbeck (talk) 11:57, 3 June 2017 (UTC)

Or you could weight the YouGov mega-poll results. Hahaha. Why not? Their dataset is massive, after all.

Consider this: if you were to drill down to the raw data for almost all of the polls published in the last two weeks, you'd find that most (seriously) have a lead of just a very few percentage points. It is only the different polling organisation weightings that have lead to the widely disparate headline figures. I am not making this up.

PS I'm not expecting you to take up my suggestion, Abso. Tho when it comes down to it, a lot of the weighted figures and the reasons for them.... being so wildly different.....the weighting assumptions are not all that far removed from wet finger and stick it in the air stuff. Boscaswell talk 18:29, 7 June 2017 (UTC)

Surveymonkey - keep or delete?

No methodology given. No data given. No information on how it selected it's sample (given it's a market research company, this may be 6000 readers of this newspaper, which would be a major bias). Not a member of any recognised polling body. Only source is a single, newspaper article in a publication with a clear political bias. In the past similar polls have been removed from these tables (see Qriously on the 2016 EU referendum polling table and the talk page discussion). I push for a swift Delete with a similar comment as per the EU ref page as to the reliability of other non-BPC polls. Jw2036 (talk) 09:48, 28 May 2017 (UTC)

• For a bit of a laugh on your bias comment which I believe shows more of your bias than SurveyMonkey, the company is commissioned by liberal bastion NBC News in the US. I'm worried a Trumpette is gonna set up a thread there in the talk page denouncing them now. I would still say Keep obviously. 88.98.211.79 (talk) 01:27, 29 May 2017 (UTC)

Delete Same reasons as above-noted. Boscaswell talk 10:26, 28 May 2017 (UTC)

Keep As well as the new SurveyMonkey poll for Scotland, this is clearly a scientific poll. SurveyMonkey is a known pollster which conducts scientific polls. In other articles on Wikipedia, SurveyMonkey polls are included. See here: Opinion_polling_on_the_Donald_Trump_administration and here: Nationwide_opinion_polling_for_the_United_States_presidential_election,_2016 There shouldn't be higher standards for UK elections. It shouldn't be necessary to be a member of the BPC. A note at polls from pollsters which are not members of the BPC can be added and is sufficient. One source is enough. That The Sun is biased is irrelevant, because that has nothing to do with the numbers of the poll. The numbers are facts. Only the interpretation of these numbers are not. Same as for Fox News polls in the US. The poll exists and should be mentioned on the polls page on Wikipedia to get the complete picture. -Ich bin es einfach (talk) 16:53, 28 May 2017 (UTC)

Keep per Ich bin es einfach. Also I don't really see how does the bias claim work here: Survey Monkey's numbers are almost the same as the average of the other polls. Rami R 16:58, 28 May 2017 (UTC)

Delete - Non BPC poll. - Galloglass 17:45, 28 May 2017 (UTC)

@Ich bin... "There shouldn't be higher standards for UK elections." There should be high standards, period. We shouldn't be restricted here by what has been done in another land on a comparable page, if we consider that to take another tack would raise our own standards or if otherwise we would not be presenting a page of an acceptably high standard. Shouldn't our policy be one of continuous improvement? Boscaswell talk 19:02, 28 May 2017 (UTC)

Keep. We do not claim that all listed polls follow BPC rules. --Wavehunter (talk) 20:09, 28 May 2017 (UTC)

• What we do require is that a poll meets the requirement for WP:Reliable Wavehunter, Surveymonkey does not, as they don't publish their data sets. - Galloglass 08:03, 29 May 2017 (UTC)
  • I think Wikipedia:Verifiability is the thing here. The results are published in the Sun so can be verified. On this page we publish the results, not the data sets. If SurveyMonkey did publish its tables and methodology, we could then be into a new debate about whether the methodology was sound. Few of us would have the expertise to do this and excluding polls because we disagree with the methodology would probably breach Wikipedia:No original research. --Wavehunter (talk) 10:53, 29 May 2017 (UTC)
    • Wavehunter I think you are missing the point regarding Surveymonkey, you do know they are not a polling company? Also that this particular survey is not a properly conducted poll, simply an online questionnaire? - Galloglass 11:28, 29 May 2017 (UTC)
      • Thank you, Galloglass, I am aware what SurveyMonkey is and I think I have understood all the points made. Are we to define what is and what isn't a properly conducted poll? The Sun says it is a poll. As per Wikipedia:Verifiability, not truth, I suggest we should only countenance deleting this poll if we can find a reliable source telling us that the poll/survey is not reliable. --Wavehunter (talk) 12:20, 29 May 2017 (UTC)

Delete. Without the normal published tables and methodology, there is no way of assessing the status of the poll. Basic things are unknown, like whether appropriate weightings were used or even whether there were steps taken to ensure that all respondents were eligible to vote in the election. Formerip (talk) 20:47, 28 May 2017 (UTC)

Keep as per previous reasonings, plus several additional facts: First, that in many other countries pollsters are not even members of any polling organisation, any such strict requirements as here laid out or anything and that doesn't make their polls invalid; second, that the article lead section clearly states that "Most of the polling companies listed are members of the British Polling Council (BPC) and abide by its disclosure rules", yet it itself leaves the way open for "other" [minoritary amount of] pollsters to not necessarily be members of BPC (and this is a wording which has been maintained throughtout several UK polling articles for years, so it looks clear that these article's essence itself is not have such a strict requirement as a limit for poll inclusion). Impru20 (talk) 20:59, 28 May 2017 (UTC)

• Delete: Data collection method too unclear for it to be considered as a reliable source. This is Paul (talk) 21:04, 28 May 2017 (UTC)
  • In countries like in Spain, for example, most polls don't even make their data collection methods clear and that doesn't make the polls invalid. "The Sun" does seem a reliable source, and there weren't any issues with it when it was the one that provided the daily trackings for YouGov back in 2010-2015. Poll results don't seem to be off those other pollsters show, and the fact that a primary source is not available doesn't make the secondary one invalid (in fact, according to Wikipedia policies it'd be the other way around, actually). One of the claims that has been made for deletion is that the publication has a political bias; well, of course it has, just show a single newspaper that shows absolutely no political bias. Yet no one seems to point that if The Sun really wanted to present a biased poll, it would show a somewhere between 10 to 20-lead for the Tories, not a lead within the single-digits that is in line with other pollsters. Yeah, The Sun publication itself may just pretend to show a big win for the Tories when their lead is clearly shrinking compared to previous weeks, but while the publication itself may be somewhat biased, we don't have anything that leads us to think that the poll numbers are. I think it's all of you who should provide sources for your actual claims and worries, actually. Impru20 (talk) 21:26, 28 May 2017 (UTC)
    • But since the general rule in the UK is that pollsters do make their data clear we should use that as a benchmark for including UK polls in this article. This is Paul (talk) 23:40, 28 May 2017 (UTC)
      • Keep The fact that a national paper with arguably the highest readership in the land uses them is enough to support the underlying methodology and their status as reputable. There are other media outlets who use them as secondary sources also which solidifies the reputation. The survey has been used widely this weekend as part of the discussion on whether the Tory lead is shrinking or whether Labour are picking up non-voters together with other polls. Let's not be blinkered here by The Sun if you don't like them. Wikipedia is not a peer-reviewed journal and as such SurveyMonkey don't need to send you an email with the raw data to analyse. The only thresholds are Reputability and not Original Research. SurveyMonkey as commissioned and published in The Sun or other papers it was in other elections is completely fine on both counts. 88.98.211.79 (talk) 01:21, 29 May 2017 (UTC)
      • Sorry, but that's not a requisite for info addition in Wikipedia. Further, using different criteria for poll inclusion depending on the country they're published would consiste bias in itself... The Sun is a notable seconday source and SurveyMonkey has been credited in other countries as a valid pollster. There isn't anything in the poll numbers suggesting that the poll is biased or that it uses a biased methodology, as has been pointed above, and such claims would require sourcing themselves once everything else is shown as sourced. The poll requirements I've read here (that, for polls to be included, their pollsters must have their data sheets published) is also entirely new for me, and ones that no Wikipedia policy or guideline I can read about do establish. Impru20 (talk) 18:04, 29 May 2017 (UTC)
        • Yes, I made a similar point just now below without having read this. No Wiki Politics project has ever had criteria for poll and polling company inclusion. Now they may do in the future and by all means anyone interested can go to the relevant portals and set up a discussion to get consensus but as it stands on current criteria, SurveyMonkey has to be included. 88.98.211.79 (talk) 18:25, 29 May 2017 (UTC)

Delete. For reason given at the top of this discussion. -- BOD -- 23:58, 28 May 2017 (UTC)

Delete as above JMaxchill (talk) 12:02, 29 May 2017 (UTC)

I cannot believe what I am reading here Editors, some of which have a clear bias towards certain publications as is evident by their comments, are arguing that if SurveyMonkey do not release the raw data and what I'm assuming patented regression model, presumably to these editors personally, they will delete a reputable source. This is absolutely crazy for the simple reason that Wikipedia is NOT a peer-reviewed journal. The only thresholds that matter is how REPUTABLE a source is and whether it includes ORIGINAL RESEARCH or is simply stating/analysing facts and/or data.

Once upon a time, I used to write Featured Articles and Featured Lists here and after having my sources reviewed I started doing my own reviews on others' nomination and I shall do the same here so that people buck up their ideas and stop peddling their personal agendas. Firstly, SurveyMonkey do NOT have to release anything to you or me to prove their reputation. Secondly, the only threshold that matters is how reputable the firm is, clearly it is the number 1 survey site in the world for both individuals and businesses, and how trusted it is by reputable media in its analysis and results. The fact that the following sources prove this beyond reasonable doubt would have led me and my fellow FA/FL reviewers to pass the SurveyMonkey source with flying colours. I repeat, Wikipedia is NOT a peer-reviewed journal.

• SM Track Record of success in polling: https://www.surveymonkey.com/blog/2016/02/29/surveymonkeys-got-polling-track-record-shines-want-proof/
• LA Times praising said track record in polling and explaining clearly why they have chosen to partner with SM for polling: http://www.latimes.com/politics/la-na-why-an-online-poll-html-20151108-htmlstory.html
• NBC using SM for their polling: http://www.nbcnews.com/storyline/data-points/poll-clinton-maintains-solid-national-lead-n667751
• 538 praising SM for accurate polling: https://fivethirtyeight.com/features/surveymonkey-was-the-other-winner-of-the-u-k-election/
• Bloomberg showing the changing trend toward webbased survey polling like SM: https://www.bloomberg.com/politics/articles/2015-11-20/surveymonkey-fills-the-gap-as-traditional-political-polls-flounder
• BusinessWire on SM Election Tracking Product: http://www.businesswire.com/news/home/20161025005674/en/SurveyMonkey-Launches-Real-Time-Polling-Map-Showcase-Latest

Really, some people need throw their blinkers away. I am not partisan in any way but an OP post clearly denouncing The Sun as biased and by extension SurveyMonkey plus other editors saying Delete as per OP is frankly shocking for Wikipedia. A lot seems to have changed since I was here and active. 88.98.211.79 (talk) 15:04, 29 May 2017 (UTC)

Delete. Get over yourself, honestly. You are missing the point entirely. If you are asserting NPOV, i.e.: 'clear bias' then you need to back that statement up. Insisting that a poll without consensus for it is posted, is itself indicative of bias on your part. You keep using the word 'clearly' and loaded terms like 'i'm sure they would agree with me' as though the more you use language like this, the more right you are, it doesn't work like that... this is not a political debating society, it's about collecting data and presenting it appropriately in appropriate categories. It should be obvious that if you put a poll of 10,000 amongst 10 polls of 100, that the larger poll is substantially different and will substantially distort any aggregate. If the methodologies are different, or not revealed, then you can't assume that they are equally valid, because there's no grounds to assert that. Stop lobbying for polls you prefer, present a scientific case for the appropriateness of a decision to include data in a set or not. Some of us are actual engineers and data scientists who take this work seriously, and sigh with dismay at people emotionalising data. 12:40, 6 June 2017 (UTC)Macdaddy

Thank you. I can't believe that we even have this discussion. -Ich bin es einfach (talk) 16:46, 29 May 2017 (UTC)

"The only thresholds that matter is how REPUTABLE a source is and whether it includes ORIGINAL RESEARCH or is simply stating/analysing facts and/or data." This is completely wrong. The issue isn't the reputation of the source, but what the criteria for inclusion in the article should be. Clearly, there should be some criteria, or we would have to include everything that calls itself a poll. The only question is whether this particular poll passes muster. Formerip (talk) 18:06, 29 May 2017 (UTC)

It clearly does. And then some. As I mentioned, the source would pass the criteria for WP:FL if this was ever nominated. If it can pass the stringest review for an encyclopedic-level article then it clearly merits inclusion in a stub list. But don't take my word for it, go over to WP:FA and ask the first editor you see who is reviewing sources in a nomination and ping them for a third party comment. I am sure they would agree with me. 88.98.211.79 (talk) 18:12, 29 May 2017 (UTC)

Also, as far as I'm aware, there are no criteria by say Wiki Politics on polls and polling company thresholds, say in the same way Wiki Music has criteria on what should be included in Album infoboxes. If you want to start that discussion on the relevant page and portal then please feel free and hopefully we can get a consensus for the future, but as it stands on current criteria, Survey Monkey passes and passed well. 88.98.211.79 (talk) 18:16, 29 May 2017 (UTC)

Delete - please note that, of course, I did not check this talk page before making my edit because I didn't even think it would be possible that rational Wikipedians could be considering keeping this survey in.

Right at the head of the article, readers are told that most polls are conducted by members of the BPC and that information is released for scrutiny. It does not say all, but as literally every other poll is conducted by a BPC member, it is unreasonable to give this one equal weighting.

Here we have the description of a poll - and not the poll itself - which allegedly took place - though we cannot see the actual questions asked. It was of 6,000 people - which is a very strange number, by the way, given the reasons for the average sample size and reeks of the Sun quitting while they were ahead. Even if we ignore the title of the source article and pretend that the source isn't biased, where are the secondary sources which lend credibility to this original source under WP:Reliability. This poll has not been widely reported - indeed, after a thorough search, it only appears to have been reported by The Sun itself - and thus does not meet WP: Scholarship.

I will make absolutely no bones about my political affiliations. If you wish to cast aspersions about whether or not I am biased, feel free. That does not change the fact that this is a straw poll.

iMarc89 (talk) 05:23, 30 May 2017 (UTC)

The Sun would be a reliable source for Wikipedia citations in Featured Articles. SurveyMonkey would be a reliable source for Wikipedia citations in Featured Articles. "Allegedly took place", "very strange number", "reeks of The Sun quitting while they were ahead" simply show you for the biased editor you seem to be without evidence and simply relying on conjecture.

Both primary and secondary sources would pass any reliability test on Wikipedia with no problems based on the weight of evidence that a number of other reputable media treat them as so as seen above. Additionally, if you said those statements in a real life public forum, both publications would be well within their legal rights to sue you for slander.

It's frankly shocking to me that the Wikipedia I edited all those years ago to create free access to information and knowledge has been reduced to slanging matches by politically driven editors. It's just sad for all involved but mostly for those who hope to rely on Wikipedia for free and unbiased information. We are not a peer-reviewed journal. Our job is simply to give the information and not start questioning national papers and the number 1 survey polling company in the world used by thousands of businesses let alone individuals. Especially when the overwhelming weight of evidence supports their reliability. 90.215.121.192 (talk) 15:05, 30 May 2017 (UTC)

Here we have an anonymous user making appeals to emotion. You haven't provided any evidence for your conjecture. I have supported my own. There are no secondary sources for this poll. The only reference to it on the internet and papers is in The Sun.

I welcomed you to cast aspersions about my neutrality, so I suppose you are entitled to do so. But your entire argument is rooted in an emotional plea regarding my and other editors bias as perceived by you. As for the "slanging match", where have I insulted any other editor in my contribution to this conversation? It is you who have dragged this conversation into the gutter by making personal attacks on me.

No one is saying that SurveyMonkey are not a reputable provider of polling services. However they themselves generally do not conduct the polls, nor are they a BPC member which removes from their gravitas in this context as their data is not available for examination - this is important because all of the other polls we have included do make their data available, thus a standard has been set. Including one, and only one, poll which doesn't meet those standards is wrong in my opinion.

What we are saying is that The Sun's opinion piece is not a reliable source. There is no mention of this poll anywhere else. It is not backed up by Reuters, the BBC or any other objective source. I stand by my comments and bow out.

iMarc89 (talk) 18:56, 30 May 2017 (UTC)

YOUR NEUTRALITY: "Allegedly took place", "very strange number", "reeks of The Sun quitting while they were ahead". Either provide concrete evidence to back up your slander or pipe down. And as I've said many times, go ask a third party from the source reviewers at WP:FA or WP:FL if either the Sun or SurveyMonkey sources would not be passed at nomination. 90.215.121.192 (talk) 19:50, 30 May 2017 (UTC)

Why don't you provide evidence for your assertions, them? Why don't you invite your third party to the conversation? My point is that a newspaper who have a clear, unhidden political agenda is not a neutral source. Nor can their research be verified as no polling data was released and their poll has not be reported elsewhere.

Nor indeed was it conducted using a sample size which meets industry standards. Sample sizes of just over 1,000 give a margin for error of precisely ±3% which is why you consistently see that size of sample. That's usually a sample of 1,000 from well over 5,000 people asked to provide balanced weightings. 6,000 is unequivocally a bizarre number because it would appear to be the sample of all respondents and would thus be unweighted. Of course, we can't test this hypothesis because, once again, we can't examine the data.

My comments are not slanderous, by the way, and The Sun have every right to prove me wrong by showing their methodology. I am saying that the source is suspect, not that it is false. I am saying that one suspect poll amongst dozens of verifiable ones is skewing information. Not all sources are equal. You know this. And to present them as such when there is clearly some statistical anomaly in their findings is misleading. The poll is an outlier and we have no means of verifying it's findings; I am justified to be suspicious.

The Opinion polling on Scottish independence article separates out suspect polls from verifiable polls so that readers can clearly see the difference between standard methodologies and those which differ from industry standards. And that's even when the polls themselves are verifiable and have been published by various secondary sources. This may be a solution here if you absolutely must include this poll. It should not, however, be included in any of the main tables.

iMarc89 (talk) 02:47, 1 June 2017 (UTC)

Keep. Anything else is censorship. It might be a good idea though to have a shading or symbol or something to show which polls do not have published data tables and/or have been adjudged by the polling industry to have questionable methodology. Dadge (talk) 19:15, 7 June 2017 (UTC)

Edit war

Perhaps it might be a good idea to semi protect the page for a while to resolve the new poll here rather than the constant edit war by anons currently going on. - Galloglass 07:47, 8 June 2017 (UTC)

PLEASE do this ASAP, @Galloglas: Boscaswell talk 07:51, 8 June 2017 (UTC)

Page protection needs to be requested - admins won't be aware it's an issue without being told. I've requested semi-protection here. 84.93.171.142 (talk) 08:19, 8 June 2017 (UTC)

if this page is going to be semi protected can someone at least put the non poll in the correct position on the table. If its to remain it must be put between Panelbase and Comres. Thanks. Obs2017 (talk) 08:09, 8 June 2017 (UTC)

protection request has now been submitted. (Someone clicked their request moments before mine) So hopefully we should have some order soon. - Galloglass 08:21, 8 June 2017 (UTC)

And it ended up with me semi-protecting the page until the polling stations have closed. Hello. I've no opinion on the inclusion of the disputed poll - it was included in the version I semi-protected, but it is still possible for established editors to edit the article. I'd encourage editors to continue to attempt to reach some consensus in the section above, or at least not continue the edit war when the protection expires. -- zzuuzz ^(talk) 08:56, 8 June 2017 (UTC)

Online? Telephone?

Whether a poll is one or the other is considered to be of great interest by many. For example, those on the website politicalbetting.com, which is one of the most widely read sites on election outcome prediction. And that fits very neatly with what this article is all about. Mike Smithson, whose site that is,writes about the use by IPSOS of telephone polling. Please leave the information there. We are, after all, a provider of information and this additional slice of information is not exactly burdensome. Thanks. Boscaswell talk 18:02, 7 June 2017 (UTC)

It's absurd to randomly report when a poll is conducted online or by telephone, and it has actually little purpose. We may be "a provider of information", but we don't provide "every bit of information"... one can just access the links (which are available) to check this out. Impru20 (talk) 18:37, 7 June 2017 (UTC)

No, we don't provide every piece of information. But we do provide information to users - our readers. Please be a little bit more civil. Who said anything about my 'randomly reporting' anything? Only you. If you were to read the article, in particular the explanations immediately above the tables, it would start to make sense to you, perhaps. I do not 'randomly report' anything. I spent my own good time yesterday evening making sure that the info about telephone or online was more complete, certainly for the past fortnight's polls, as another editor had added that information for one pair of Survation polls but no others. The information I had presented hanged together very well, until you unilaterally deleted what there was in the table, before giving me a chance to respond here to your angry and rude comment. Boscaswell talk 19:18, 7 June 2017 (UTC)

Please, may I question where I was uncivil? Only me said it because there hasn't been much time for these to be shown to people. It was "random" because it was applied to just a few Survation polls and a Ipsos MORI one. The text above the tables says that all polls are online unless otherwise stated, yet you also noted online Survation polls. Two sets of those Survation polls were consecutive; another one was separated from the rest, and then that lone Ipsos MORI poll. And that's it, no other polls. I'll say that it looked random, yeah, so I dunno where the uncivility is. I'll also ask you to be much more civil (not just a little bit) because I've not insulted you and now I have to read you depicting my comments as "angry and rude". Not needed at all.

Anyway, I think your issue may be solved with several footnotes, which (I believe) are up for use precisely for these kind of things. Impru20 (talk) 19:33, 7 June 2017 (UTC)

Instead of trying to annotate every poll in the table, what about a paragraph of text explaining that the pollsters use different methods with some examples of who uses what plus links to discussions of methods? Bondegezou (talk) 09:41, 8 June 2017 (UTC)