Misplaced Pages

This is an old revision of this page, as edited by Rami R (talk | contribs) at 17:08, 14 June 2020 (→Use of Bristol Elects constituency polls: [[Talk:Opinion polling for the 2017 United Kingdom general election/Archive 4BPC membership not previously required). The present address (URL) is a permanent link to this revision, which may differ significantly from the current revision.

Elections and Referendums Unassessed

This article is within the scope of WikiProject Elections and Referendums, an ongoing effort to improve the quality of, expand upon and create new articles relating to elections, electoral reform and other aspects of democratic decision-making. For more information, visit our project page.Elections and ReferendumsWikipedia:WikiProject Elections and ReferendumsTemplate:WikiProject Elections and ReferendumsElections and Referendums ??? This article has not yet received a rating on Misplaced Pages's content assessment scale.

Politics of the United Kingdom Unassessed Mid‑importance

This article is within the scope of WikiProject Politics of the United Kingdom, a collaborative effort to improve the coverage of Politics of the United Kingdom on Misplaced Pages. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.Politics of the United KingdomWikipedia:WikiProject Politics of the United KingdomTemplate:WikiProject Politics of the United KingdomPolitics of the United Kingdom ??? This article has not yet received a rating on Misplaced Pages's content assessment scale. Mid This article has been rated as Mid-importance on the project's importance scale.

Re-order columns?

We currently have: SNP, Green, BXP, Plaid. But in previous articles we've had Plaid after the SNP, being the 2 nationalist parties, for Scotland and Wales. I feel it would be far better to have the Brexit Party last, as they will inevitably drop out altogether. Boscaswell talk 09:16, 14 December 2019 (UTC)

PC and SNP should be together because historically some pollseters group the separatist vote. May happen again. As for the Brexit party, they could stay for the 2020 table, they have been a significant party in 2019 Jonjonjohny (talk) 09:30, 14 December 2019 (UTC)

Support pooling SNP and PC. In any event they should be next to each other.Cutler (talk) 13:18, 14 December 2019 (UTC)

Strongly Oppose pooling SNP and PC. Almost all pollsters now poll them separately, including those who had 'SNP+PC' lines in the past. See YouGov, Survation, BMG, ComRes, Ipsos MORI, etc... And those who don't poll PC separately pool them with 'others', NOT with the SNP. So separate data IS available, would be daft to pool it, and 'may happen again' is just about the lamest excuse for a justification. Besides they are separate entities with both its own representation in Commons. Would you pool SF and SDLP in the NI polls because both are 'separatists'? Or have a 'SNP+PC' column in the Scotland and Wales polling? This looks again like a politically motivated attempt to obfuscate by misrepresentation. JezEuansson (talk) 16:37, 23 December 2019 (UTC)

Let's wait and see. If no opinion polls show SNP and PC together, then the columns can be separated according to their 2019 results. If some opinion polls resort to their traditional method of polling them together, then their columns should be together. Let's not reach premature conclusions before we even have the actual data. Impru20 23:27, 23 December 2019 (UTC)

Strongly Oppose pooling SNP and Plaid. It’s rare for pollsters to combine them anyway. Boscaswell talk 10:36, 24 December 2019 (UTC)

Strongly Support pooling Conservatives and Brexit Party because they are both English nationalist hardline Brexiteers. Makes just as much sense, if not more, as pooling SNP and PC because they are 'separatists'. JezEuansson (talk) 10:56, 24 December 2019 (UTC)

• Support. I agree with Boscaswell that if any of these four parties is going to be removed from the table, it is the Brexit Party, considering the party's current downward trend. The other parties you've mentioned are relatively stable in polling averages. However, I am opposed to combining any of these parties in any way, since Plaid Cymru and SNP are clearly separate parties with different goals: one is for Wales and the other is for Scotland. --Comment by Selfie City (talk about my contributions) 12:51, 24 December 2019 (UTC)

The issue comes because some pollsters use(d) to combine SNP and PC into a single figure for both of them. That's the only purpose of them being combined, casting aside the ideological nature of some comments being posted above and which do not belong to this discussion. If pollsters do not show a joint figure for both of them any longer, the need for such a combination vanishes, obviously. Impru20 12:56, 24 December 2019 (UTC)

Correct. Wales and Scotland aren't even near each other. --Comment by Selfie City (talk about my contributions) 14:42, 24 December 2019 (UTC)

Comment The opposes seem to be blurring the debate between 'pooling' the nationalists & putting the columns next to each other. They are separate topics. Given the possiblity of a pollster (even a rare one) lumping them together, not putting them next to each other, so they could be joined for that poll, seems extremely foolish. 86.175.154.209 (talk) 23:02, 1 January 2020 (UTC)

Complaint Who did approve removing Plaid Cymru from the table? Of course nobody did and it should be reinstated at once. And by the way when you vandalize at least do it right as the total for the BMG poll is now 102.5%. Daft. JezEuansson (talk) 10:09, 19 January 2020 (UTC)

Split Pollster/client into separate columns

I propose we split Pollster/client into separate columns, which has the significant advantage that you can sort on pollster retaining date order, to see how that pollster-sample changes over time easily (or whatever the previous sort was). Also allows sorting by client (newspaper etc) if that interests the reader. Currently sorting on the column is largely by pollster but split into client subgroups so not a continuous pollster-sample-method by time (or previous sort). This was suggested for the 2019 article but would have been a lot of work, whereas now is the easy time to make this change. Rwendland (talk) 14:12, 15 December 2019 (UTC)

This seems sensible to me. Something like this?

Tangentially, I'd also be happy to see Client(s) changed to Client and Date(s) conducted to Dates conducted. (There are polls that only have a single day's fieldwork, but they're too rare to justify the (s), aren't they?) Ralbegen (talk) 14:46, 15 December 2019 (UTC)

Mild oppose - fewer columns the better - doesn't add any extra information - not really that interesting in any event.Cutler (talk) 19:42, 15 December 2019 (UTC)

Maybe I'm unusual, but when different poll results are a bit erratic as during last election period, I was very interested in seeing each pollster in time order to see the direction of change for each pollster. This was on the assumption their different sampling/correction methods were affecting prediction, so seeing a single method timeline might be more reliable show change direction. Rwendland (talk) 10:15, 16 December 2019 (UTC)

Mild Support or mostly harmless. I think it will be not really be anymore challenging for readers or editors and does allow interested readers to simply see how individual pollsters and/or clients change. ~ BOD ~ 10:31, 16 December 2019 (UTC)

As the above shows support, if mild, for this suggestion (3 support, 1 oppose) I plan to implement it in a few days time, per Ralbegen version above, unless there is strong objection. Rwendland (talk) 15:56, 18 January 2020 (UTC)

Mild Support as above. Bellowhead678 (talk) 16:24, 18 January 2020 (UTC)

Mild oppose I don't think the client makes enough of a difference to polling to be listed as a separate column, but if it allows all polls from a specific pollsters to be ordered correctly then I'm not worried. I would prefer to have the dates listed first in the table however. Clyde1998 (talk) 17:28, 3 February 2020 (UTC)

Graph

Following SelfieCity's suggestion, I've changed the graph from Nationwide opinion polling for the 2020 Democratic Party presidential primaries so that anyone can change the graph. Here it is with the UK names and colours but the US numbers left in to give an example of what it might look like. It should be fairly self-explanatory for others to add new polls, just add the date to the end of the list of x-values and add the Conservative score to the end of the list of y1 values, the Lab score to the end of y2 list and so on. Bellowhead678 (talk) 21:49, 15 December 2019 (UTC)

Looks great! --Comment by Selfie City (talk about my contributions) 18:01, 23 December 2019 (UTC)

I'd be very happy having that - does mean that no-one is having to constantly update an image file and is very clear. Also could be copied for use in the sub-national polling section or for seat estimates moving forward. Clyde1998 (talk) 05:43, 27 January 2020 (UTC)

Graph is fine, but the bar chart on the right of the chart on the article doesn't sense check: it is comparing UK election numbers with GB polling numbers, and also the averages are not the averages from the numbers in the tables. I think the bar chart should compare an average of GB polls (which most are) with the GB numbers from the election, and the arithmetic needs to be checked. Also, the coding below looks like a brute to maintain. RERTwiki (talk) 11:48, 30 January 2020 (UTC)

Agree with the comment above. Election 19 values should be the GB result, not UK. Also the way the current vote shares are calculated should be explained as it is obviously neither the arithmetic average nor the weighted average of polls in the table. JezEuansson (talk) 11:53, 30 January 2020 (UTC)

Graphical summary

Graphs are unavailable due to technical issues. Updates on reimplementing the Graph extension, which will be known as the Chart extension, can be found on Phabricator and on MediaWiki.org.

WARNING There's something seriously wrong with this graph. The SNP never had 25% support and the Brexit Party didn't die in March. Boscaswell talk 22:38, 23 December 2019 (UTC)

As I said, I left the US values in to show what it looked like. Bellowhead678 (talk) 09:13, 24 December 2019 (UTC)

Gotcha, Bellowhead678! Boscaswell talk 22:23, 24 December 2019 (UTC)

@Bellowhead678: Should we move this into a template and then we can copy this into the article about the election itself? Also. I feel like there should be some horizontal lines in the graph. it's pretty difficult to see the percentages. if there was a horizontal line every 10% it'd be a lot easier to read the graph!--OskarsC (talk) 11:08, 18 February 2020 (UTC)

I'd support both adding horizontal lines (although they shouldn't be too dark as it distracts from the lines) and moving it into a template. However I have no idea how to do either of those! Bellowhead678 (talk) 18:28, 18 February 2020 (UTC)

Like this. You can edit the template by just searching in the searchbox "Template:Opinion polling for the next United Kingdom general election". Adding the horizontal and vertical lines was really easy. All you had to do was look up "Template:Graph" and there is a detailed tutorial on how to use it.--OskarsC (talk) 09:10, 23 February 2020 (UTC)

In case needed

I've just deleted these from the article until such time as or if we get polling for these regions. Archiving here for easy re-insertion.

London

North East England

North West England

Yorkshire and the Humber

East Midlands

Shelving these too, if it's okay. There aren't any Scotland/Northern Ireland/Wales polls yet either. (And as NI only has the one pollster, it'll probably be some time before we get any for there). Ralbegen (talk) 15:41, 23 December 2019 (UTC)

Ralbegen, do you have the number of votes for each of the above, to go into the Sample size fields? Boscaswell talk 04:04, 5 January 2020 (UTC)

I didn't write them out, sorry, I just archived the below tables here on the Talk page (after Bondegezou archived the above tables) so that they weren't sitting stranded in the mainspace. Though I note that the sample size field for the general election nationally has been removed seeing as an election is different to an opinion poll, which I suspect is the better way forward. Ralbegen (talk) 10:09, 5 January 2020 (UTC)

That's a shame, Ralbegen. I always thought that the number of votes is useful & interesting info to have there. But I'm not about to get into a disagreement with the editor who took it out. Some editors can be more belligerently argumentative than others. ;-) Boscaswell talk 01:20, 6 January 2020 (UTC)

Scotland

Wales

Northern Ireland

Article's purpose

There are no opinion polls yet and the only edits seem focused on a rather weird struggle over how many regional/local results get to be included into the article. I think it is obvious that opinion polling will start coming out in shortly (hopefully), so I don't think deleting the article will prove useful at this point, but the actual issue is: What is the current purpose of the article aside of being a placeholder-in-waiting for opinion polls to come? Shouldn't it have been hold off until there was at least one opinion poll to list? Impru20 15:48, 23 December 2019 (UTC)

It's going to be pretty bare for a bit. It might have been better to have kept it in draftspace until there were any polls, but there are a lot of eyes on this article series, so its premature start was probably inevitable. Ralbegen (talk) 16:33, 23 December 2019 (UTC)

No, leave it there. And it should always be there, even before polls happen. This way, little problems can be ironed out, such as column order. What's the point of hiding it only to bring it back a few days later? Doh! And in any case pollsters are late this year. Usually, 12 days from polling day, there's been an opinion poll. Boscaswell talk 22:36, 23 December 2019 (UTC)

Maybe you are forgotting what Misplaced Pages is and what it is not and the Misplaced Pages's notability criteria for article creation. Any other article being created before the information it intends to cover is even reported by reliable sources would be deleted outright, because it would fail the primary criterion justifying its very existence: verifiability. For us to attempt to presume when and how will opinion polls start coming out would be a violation of WP:CRYSTALBALL. The current article lacks any sources on its topic, and none can't be provided because no opinion poll has been published yet. So yes, it's premature (Doh!).

Further, opinion polling articles exist because of WP:SPLIT, i.e. the amount of information they typically handle being too large for the main article to cover. No information excess, no separate article (much less if no information at all exists yet!). These are basic Misplaced Pages rules that the premature creation of this article, out of some desire for rushing it into existence, is breaching.

I think we can be fairly sure that opinion polls will eventually come out. However, the article currently serves no purpose other than for some people to start some preparations, then for other people to remove such preparations, while they kept clashing on who's the better one at predicting the territorial scope of the opinion polls that will come out next. And I'm sorry to point out the obvious, but Misplaced Pages is not for this nor should we be wasting time and effort on a topic which does not exist yet. Impru20 23:19, 23 December 2019 (UTC)

While it may have been better not to create the article until the time was right, what's the point in moving the article back and forth between mainspace and draftspace, requiring an administrator? Now we've got it prepared, we might as well keep it here until a poll comes out, at which time we can add the poll to the article and have no concerns over complex page moves. Why create unnecessary work for ourselves? --Comment by Selfie City (talk about my contributions) 13:13, 5 January 2020 (UTC)

Update tag

An update tag has been added to the article, but I'm not sure anything needs to be updated. As far as I'm aware, there haven't been any polls since the election. Bellowhead678 (talk) 09:28, 13 January 2020 (UTC)

Agreed. Let's remove. Bondegezou (talk) 09:40, 13 January 2020 (UTC)

Plaid Cymru

Although historically it’s always had its own column, an editor has gone and removed that. Without first taking it to talk. I think it needs to be there, since although there’s always a small percentage, it’s highly unlikely to fade out completely and more importantly, the party has 4 MP’s! Boscaswell talk 23:19, 18 January 2020 (UTC)

In theory, we should reflect how reliable sources report the polls. So far, BMG headline the Conservatives, Labour, Liberal Democrats, the Greens, the Brexit Party and Ukip. Opinium includes the same but adds the SNP and Plaid Cymru. I can't find any media write-ups of either poll yet. So on the level of sources it's inconclusive at the moment. I maintain my view from the page for the previous election that including the SNP and PC in the GB/UK poll table is unhelpful because the changes in their poll scores tell the reader absolutely nothing about how well the parties are performing because: all the changes are by necessity within the margin of error, and even if there were changes dramatic enough to be reflected in national poll results, they would be much clearly captured in polls of Scotland and Wales... But we should wait until there's a clear pattern in reliable sources that we can follow. Until then I'd support restoring Plaid Cymru to the table. Ralbegen (talk) 00:52, 19 January 2020 (UTC)

I didn't think it would be controversial but their results are very low and not considered notable by reliable sources. I don't see any reason to give Plaid Cymru its own column here. Onetwothreeip (talk) 01:29, 19 January 2020 (UTC)

IIRC, the main reason for including Plaid previously was that YouGov had a single SNP/Plaid option prior to the European elections in 2019. Since then they've separated them from each other. Realistically, Plaid are going to be on either 0% or 1% because of the size of Wales and Plaid usually being between 8-15%. However, it's still useful from a trends point of view, as people can work out averages from the data - with more 1% results suggesting Plaid are doing better than if they were getting more 0% results. I'd say we should keep them for now as most pollsters include them in their initial prompt (at least for voters in Wales), which should really be our guideline as for who's included: are the party prompted by multiple pollsters? If the answer if yes, regardless of their territorial extent, then they should be included in the table. Albeit, I'm be happy what's being used right now where named others in the poll (eg. Plaid, UKIP, Change UK) are added as a note next to the 'other parties' percentage to keep the table reduced to size, whilst allowing people who want the polling data for smaller parties to find it easily. Clyde1998 (talk) 05:39, 27 January 2020 (UTC)

I strongly suggest Onetwothreeip stops vandalizing the results table and removing Plaid Cymru. Discussion makes it cleat he is the only one wanting to remove them and has absolutely no support from others. As mentioned above, removing a party that is regularly prompted by all pollsters is total illegitimate and should be explicitly forbidden. JezEuansson (talk) 12:08, 16 February 2020 (UTC)

I would support removing Plaid Cymru given their very low scores - if they rise again we can always reinsert them. Bellowhead678 (talk) 09:51, 17 February 2020 (UTC)

Not the issue Plaid Cymru have MPs which is already one good reason NOT to exclude them. And now all pollsters including YouGov have a separate entry for PC in their published results, you just have to check the sources. Which is a second reason NOT to exclude them. Clearly there is absolutely NO consensus to remove them, ergo they should stay. JezEuansson (talk) 10:59, 17 February 2020 (UTC)

The consensus here is clearly to keep them in the table but to merge their results into the Other column, with a note for the Plaid Cymru results. Plaid Cymru have only 4 MPs out of 650, and don't poll above 1%. Onetwothreeip (talk) 08:39, 19 February 2020 (UTC)

There is absolutely NO consensus here, only you want to remove PC. Now this footnote idea is the daftest possible. There were 468 polls between the 2017 and 2019 elections. So what you propose is polluting the page with 468 footnotes only because you have a fixation on PC? Get a life, lad and stop vandalizing. JezEuansson (talk) 21:40, 22 February 2020 (UTC)

Graphical Summary

Tories needs to be changed to Conservatives. There's enough space. No nicknames or colloquialisms.  — Calvin999 13:27, 30 January 2020 (UTC) Fixed! DerÖsterreicher1 (talk) 22:18, 30 January 2020 (UTC)

Graphical polling

I think, we should go with my version, due it is like Avopeas' Version of his works (Iceland, Austria, Denmark, Norway, Sweden and Finland).these are very popular in my opinion. But what do you think? DerÖsterreicher1 (talk) —Preceding undated comment added 00:25, 31 January 2020 (UTC)

I don't see the value of a graph that is mostly blank space. It makes more sense for the graph to go up the most recent polls. Bondegezou (talk) 09:57, 31 January 2020 (UTC)

I am more concerned by the fact that the graph shows a made-up average and compares it with the election result. Is such an average backed by a reliable source? One thing is to calculate average trend lines, a very different one is to show specific, hand-made results. I also don't see these graphs as "popular" (yeah, one user has added them all over Misplaced Pages but I don't think that counts in terms of popularity). In fact, these have been removed/replaced from articles when an alternative version has been made available. Impru20 10:06, 31 January 2020 (UTC)

Both graphs use averages for smoothing. I'd personally prefer a handmade graph like DerÖsterreicher1's since it include the actual polls as data points. I also don't see the problem with the blank space. It's only been a few weeks since the election, why would you want to try to hide that fact by stretching the timeline? --Gbuvn (talk) 11:49, 31 January 2020 (UTC)

Using averages for smoothing is one thing and well into WP:CALC. Using a specific average as indication of something, then compare it with election results and present it as such, could go into WP:SYNTH territory in my opinion. Impru20 12:10, 31 January 2020 (UTC)

I concur with Impru20 about the bar charts on the right: these are straying further into WP:OR and should be avoided. Bondegezou (talk) 13:24, 31 January 2020 (UTC)

Technically, the line chart also compares a specific current average (and former averages) with the election result, it's just harder to see the exact numbers. But that's not my point, I don't really care much for the bar chart. I'm just saying that the graph template is less customizable, shows less information and can get quite ugly, which is why we should use an image as long as long as somebody is willing to update regularly. --Gbuvn (talk) 13:45, 31 January 2020 (UTC)

I am ready to update it regularly, but you see, many people are against it, even though i made this Graphic even before the first poll (with the results of the general election only, then i added the polls, had one mistake on it, and it got replaced by this ugly chart, i agree, but I have here much Opposition, and as long this is a thing, I can't really do something. Because my edits will be reverted twice, before i can make my edits! And i do not want a Wiki War + Ban so....DerÖsterreicher1 (talk) 14:41, 31 January 2020 (UTC)

Personally, I much prefer DerÖsterreicher1 version of the graph, its much clearer. All these graphs have used averages, and the blank space is better than uninformative pointless parallel lines. ~ BOD ~ 15:09, 31 January 2020 (UTC)

It is not the fact of using averages to calculate trend lines. It is the fact that one of such averages is cherry-picked and presented as the headline figures in comparison to election results. Why those and not any other? This is very obviously original reseach, unless there is some source backing up the use of such specific average as a headline figure. Plus, in most charts of this kind averages tend to be moving ones, which make the presentation of any specific, "headline" figure as pointless as these just grasp a very limited point in the trend which, by itself, does not have to be representative the rest of it.

The blank space is a non-issue for me, though. Impru20 16:07, 31 January 2020 (UTC)

Yea so, Problem is, you are the only one for this Ugly (easy manipulatable) Graph, that was made there. Well, If you like easy manipulatable Graphs, fine by me, what my problem is with that, is, that I (and mostly I think others too) hate manipulation. With an Excel Graph, you can't let anyone edit it, and then manipulate, that is why I am for regularly-updating Users, who make a Graph for an Opinion Polling page. But, what I have a problem too, that you involve into things, and then bring Arguments, that have nothing to do with the other things, for example, when I wanted to add my Graph to the Opinion_polling_for_the_next_Spanish_general_election, I added it first, you deleted it, for almost no reason, except for "Consistency", which is a "very logical explanation". No that's no proper argument, it's just, you say , it's a verry logical explanation, and idk what was explained, please explain it next time properly, because one word, and then it's a very logical explanation, No, I don't get it, I am sorry, that is Harvard-Like Thinking. But hey, I didn't open my mouth, I was happy with your decision, at hte greek Opinion Polling, Ok, I involved into it, even tho there was a graph, wasn't cool for me. But here I draw the line, I had a graph done before even the first Poll came out, Updated it ASAP, when a Poll came out, Uploaded it yesterday with a Mistake, calling the Conservatives Tories, and you instantly stay ith hte manipulatable graph. let me guess, Consistency, right?DerÖsterreicher1 (talk) 16:26, 31 January 2020 (UTC)

First, this "easily manipulatable graph" was the result of a consensus in a discussion above, one you just rejected to abide to by trying to enforce your chart without a previous discussion on it. Surely your current tone is entirely out of place here.

Second, other users have voiced opposition to your graph.

Third, why is the current graph "easily manipulatable" but yours isn't? In fact, yours is presenting a random, cherry-picked average as headline figures to casual readers, whereas the other doesn't, causing WP:OR concerns. You are not even addressing this issue.

Fourth, if you have a personal issue with me because I did not agree with your unilateral addition of your graph to Opinion_polling_for_the_next_Spanish_general_election that's your problem, not mine. Nonetheless, just note that this one you point out is just one out of many examples of you seeking to replace other charts with your own all throughout Misplaced Pages in an unilateral way and without seeking any consensus. I am pretty sure that "consistency" is a much more logical and policy-based explanation that "hey dude I like my chart let's have it instead of the other one whose model has been used for years in Spanish opinion polling articles yeah?"-sort of explanation. What'd you think if I just started uploading charts for every opinion polling article where you have had your charts imposed, then unilaterally sought to replace them all? You would surely complain in a similar way as you do now, right?

Fifth, yeah, when someone disagrees with you they may "bring arguments". It's a pity that you have a problem with that, because that is how Misplaced Pages and consensus-building works.

In my opinion, I would rather have this consistent and use a chart similar to the one used in Opinion polling for the 2019 United Kingdom general election. Though at this point there are just three opinion polls published, which may not even merit a chart yet. Impru20 16:41, 31 January 2020 (UTC)

But is DerÖsterreicher1's graph that much different from the 2019 one? I understand that you're opposed to the bar chart part but what about the rest? I feel like it should be possible for everyone here to agree on using that graph without the bar chart. --Gbuvn (talk) 20:04, 31 January 2020 (UTC)

In my opinion, there are a number of issues where the 2019 chart is clearly an improvement: poll dots being much more visible (if you choose to have them, I understand that it is for them to be spotted), grid should be less visible (it is the trendlines that should feature predominantly; a very notorious grid will only serve to mislead), unneeded decimals in the Y axis... such things.

Aside of the fact that three polls is not even enough for a comprehensive chart to be created, but that's a different story. Impru20 20:20, 31 January 2020 (UTC)

You know what? If you really do not appreciate my work, fine by me, I will not do it, case closed, thanks Impru, I will focus on the other things then more, I just wanted to do goo work on this Misplaced Pages, but if you can't appreciate it, then fine by me, keep this easy manipulable Chart, if possible until 2024. Same with Spain, after you reverted it, I did not care anymore, so is it now with the UK, thanks man, you gave me less work! Even tho I like, no I LOVE the UK, I will not continue this work because of Impru's Negative attitude to my work! DerÖsterreicher1 (talk) 20:50, 31 January 2020 (UTC)

Oh no no no. One, I am free to give my opinion on your work, just as I have spent almost nine years in Misplaced Pages receiving opinions on my own work as well. This is a collaborative project, and it's not meant that your work must be automatically accepted or liked by everyone; consensus-building and BRD exist precisely for this. Second, you complain about others' "negative attitude to work" while concurrently showing a complete disregard to the current chart (i.e. other's work) by repeteadly dubbing it "easy manipulable". Yes, it is far from perfect, but it had received at least partial acceptance in this talk page before you unilaterally added yours. If there is any willingness in bringing this discussion to a constructive outcome, I'll be all for it. If this is to be limited to a "accept this or I will leave this forever"-issue, then I would not be engaging in such drama. Impru20 21:28, 31 January 2020 (UTC)

Sorry I missed this discussion. I'll just give some background to the above discussion. For the 2017 election, I kept the graph updated every few days. After I was away for a while, someone else created an alternative graph and was happy to update it every few days until they were also away for a couple of weeks/lost interest and I recreated my graph. It's perfectly possible to do this again, it just means that it's 1. difficult to check 2. if someone is away then the graph cannot be updated. (However, it does make it more difficult to vandalise) Someone therefore suggested using the model used in the US primary polling which anyone can update. I agreed that it's too soon for a graph, but seeing as there was one already, I thought we should use the graph agreed earlier and which anyone can edit. The average is calculated by taking the mean of the previous polls (I suggest ten but we can agree that later). If we make a single point for one day (rather than one point for each poll even if there are multiple polls on a single day), then it does avoid the jerky lines from earlier graphs. Bellowhead678 (talk) 12:53, 3 February 2020 (UTC)

Not invested in the bar chart thing but the points for individual polls should be displayed as in previous elections regardless of how the curve is calculated I think. —Nizolan 13:11, 3 February 2020 (UTC)

I don't think there's a way to do that with the graphs generated directly on Misplaced Pages. Bellowhead678 (talk) 13:39, 3 February 2020 (UTC)

They do it for Canadian federal election articles and I think the method is a similar one. Impru20 13:52, 3 February 2020 (UTC)

That graph is created in R and looks great. The point was, that the built-in template doesn't allow adding data points, which is why I would not use it. --Gbuvn (talk) 13:57, 3 February 2020 (UTC)

Fair enough, though just IMO that might be a serious enough deficiency in the Graph extension to recommend using a graph made with an external programme instead. I'm not too sure about presenting Misplaced Pages's own mathematical analysis without a graphical indication of the original data. —Nizolan 14:17, 3 February 2020 (UTC)

Folks - this conversation seems to have run out of steam without causing the article to reflect collective views. I'm sure this will be controversial, but I would suggest that criteria for a 'good' chart are:

1) Each individual datapoint, as listed in the table in the article, is visible on the chart 2) The trend-line is an easily calculable (checkable) function of some well-understood subset of the datapoints 3) The trendline is 'stable': once the trendline on a given day is established, it will not change as a result of datapoints added in the future 4) The trendlines are the most visible elements of the chart, but the datapoints are visible enought give a natural indication of the dispersion of the data 5) Ideally the sample size of polls is indicated on the chart (datapoint size) 6) ideally the date associated with each datapoint is the mid-point of its survey period 7) There should only be a single trendline point at a given time (no 'jags')

The current chart either does not meet some of these criteria, or else it is entirely opaque (to me) whether it does.

I'm sure I can create a few alternatives in Excel for people to choose between, and I would be happy to update one for each poll at the moment. However, something smarter which would let anybody make updates would be better, especially in the next campaign period when the updates could turn into a full-time job. I'm afraid I don't know the technology well enough to make any suggestions, but I'd be willing to give a hand if some setup were needed.RERTwiki (talk) 09:36, 27 March 2020 (UTC)

Your criteria seem on point to me. I'd agree - there's a conflict between having a good quality graph and updatability. One thing that was tried the last time was having an open Google sheets document and a workflow description for anyone to update, however it seems to have been too complicated for most users since it included post-processing in Inkscape. --Gbuvn (talk) 10:18, 27 March 2020 (UTC)

Anybody give me a clue where to find - even how to look for - the code/data for the current chart? I see the code in the source seems to be a reference to the page itself. I'm clearly missing something, but don't really know where to begin to look for this info. RERTwiki (talk) 10:20, 1 April 2020 (UTC)

OK, I'm catching up a little. I found the template for the chart.

The template has the sole advantage that it is 'open' for editing, but is otherwise poor. The list of values is not maintainable, and the only way I can see to catch and fix errors is to independently re-calculate the values. The charts have very poor features compared to Excel: most of the requirements I list above can't be met, as far as I can see. Not only that, the line on the chart is based on an unspecified moving average, with an unspecified 'cubic interpolation' on top of that. Am I the only one who finds this whole situation very unsatisfactory? RERTwiki (talk) 11:27, 1 April 2020 (UTC)

In Nationwide opinion polling for the 2020 Democratic Party presidential primaries, it’s easy to calculate because we use averages of polling aggregators to achieve our own numbers to insert into the graph. For example, if Joe Biden has 50% according to one website, 46% according to another and 48% according to the last one, we input 48% into the graph as his average until the polling aggregates change. --Comment by Selfie City (talk about my contributions) 12:23, 1 April 2020 (UTC)

Thanks for the response. There are two problems:

1) if you made a typo and input 47% instead of 48%, how does the error get caught? These numbers are supposed to be averages of - well actually, who knows how many polls. In the UK there are many parties - about 7 - not 2. that's a multiple of the chance of error. Can you imagine what the template is going to look like when there are literally hundreds of datapoints in each of 7 comma separated lists? Maybe 70 datapoints might need to be changed if someone discovers a poll that was missed. You have to pick those points from the lists by dead reckoning. That's an invitation to a trainwreck.

2) The graph format is absolutely pants. I've thought about some heroic coding to update or create a new template, but I would be trying to re-create significant features of Excel, after I fought my way through learning whatever language it's written in. That's major re-invention of the wheel, and doesn't sound like the right way to go, especially given problem 1)

I think the way to do this is to have some shared empty Excel sheet which anyone can download, paste in the current table of polls from the wiki page, and run a macro to create the chart. Unless we can use a native Excel output format such as PDF, there is a problem in creating the output graphics format. I've used PDF2PNG before, but in my experience that isn't very reliable.

Any comments? Any experienced editors know how this is best done elsewhere in Misplaced Pages? RERTwiki (talk) 09:30, 3 April 2020 (UTC)

I missed this comment. I'll reply to you today when I have the time. --Comment by Selfie City (talk about my contributions) 11:39, 10 April 2020 (UTC)

RERTWiki, I’m starting to notice unusual curves in the graph that lead me to believe that it could be much better. However, I’m sure there are people who can do the necessary code to improve Misplaced Pages’s graphs in general.

However, I believe that whatever graph is used, like this one, should be designed so that anyone can edit it directly on Misplaced Pages or at some other open-source location. If you know anyone who could do the code, I think we should give them an opportunity to try improving or making a better graph. --Comment by Selfie City (talk about my contributions) 00:57, 11 April 2020 (UTC)

It looks like the issue with the lines has now been fixed. --Comment by Selfie City (talk about my contributions) 17:23, 11 April 2020 (UTC)

Thanks for the reply, just saw it now. Yes, the lines were crazy and are now 'fixed', but what they are is still desperately obscure. I don't think we should sacrifice a clear presentation while waiting for the graphic capabilities of Misplaced Pages charts to catch up. If that capability existed, it would be better to use it. But it is a huge task, and currently the graphics are terrible. We can do much better by simply sharing an excel file (somehow) and letting anyone with access to Excel (millions of people) update it.

Second, here is an example of what we could do. This is a chart of the last 10 polls, averaged as discussed above, in as far as I can quickly mock up in an hour or so. RERTwiki (talk) 16:07, 17 April 2020 (UTC)

I think the Microsoft Excel version is the best alternative of those that have been suggested so far. However, instructions should be added near the graph (presumably in the form of a comment, making it invisible outside of editing mode) to let editors know how the graph can be edited and posted in the new version.

I would suggest developing a Microsoft Excel graph and then building a consensus to use it over the current graph. As long as it's not too hard to edit (not much harder, if harder at all, than the current version, would be my qualification), I would support that change, as long as it was superior in design to the current graph (which it sounds like it will be). --Comment by Selfie City (talk about my contributions) 22:37, 17 April 2020 (UTC)

OK: I've modified the Excel so that it will now update easily. I've put the instructions on a page in the spreadsheet, and this is what they say:

To Prepare the Chart of Polling Averages

1. Copy all the cells from the Misplaced Pages table of polling data: select them and press control-c

2. Goto the sheet in this workbook called 'Table': select cell A1 and press control-v

3. The chart on the page 'Averages Chart' should now be up to date.

4. Select that page. If it looks reasonable, from the 'File' menu select 'Save As' and save the chart as a PDF.

5. Upload the PDF to Misplaced Pages and insert into the page as required

Possible Errors

1. Averaging the last 10 polls in the table makes most sense if they are the last 10 collected. Column AB in 'Table' highlights in pink polls which might have been inserted in the table out of order. (Order published might not be mid-point order) The option here is to fix the table order if you feel the error is significant.

2. If the data entered into the table for Con, Lab, Lib, SNP, Plaid and Green are not numbers (eg "3%"), the datapoint charted will be zero, and the averages will be wrong. The data should really be fixed.

3. The date formats get confused. The most complex part of the sheet is parsing the column of start and end-dates and translating the text into two columns in Excel date format. If someone puts something unexpected in the dates column, it may break the chart.

4. If there are more than about 1000 polls, and the earliest polls and averages will be omitted. In which case the spreadsheet will need a minor edit to extend it down and extend the chart to cover the new data.

Limitations

1. Currently Brexit seems to be aggregated with Other in some polls. Brexit and Other are both excluded from the chart to ensure that the data displayed is reliable. If someone wanted to fix the table it could be extended to include Brexit.

I'd welcome comments, and any suggestions as to where to put the sheet so that everyone can access it. The Spreadsheet contains no Macros, since I know there is a reasonable concern about viruses. We could have a better chart with Macros, specifically individual datapoints could be customised. Again I'd welcome comments on whether that would be worth it.RERTwiki (talk) 09:13, 20 April 2020 (UTC)

IMO, only a few parties should be included: Labour, Conservative, LibDem, SNP, and Green. As long as those are included, we don't need to worry about the others, as UKIP and the Brexit Party have low polling scores and are largely irrelevant at the moment. This sounds good, so assuming we can gather consensus for the new graph here, I would support adding it. I would personally count consensus as only a few people as not much more than that is realistic. --Comment by Selfie City (talk about my contributions) 13:35, 21 April 2020 (UTC)

Alright, we can wait a day or so to see if there is any more feedback, and give it a try. Meanwhile, here is a link to the sheet in dropbox:

dropbox link

When I click the link there are buttons on the right which give me options to download the file. If you could test that and it works, we might be converging on a solution. RERTwiki (talk) 09:05, 22 April 2020 (UTC)

The best graphs we've had on this article series was Mélencron's LOESS graph:

Unfortunately they've retired now, but if we want to move on from the inline one, that's what we should be aiming for. Excel isn't appropriate for generating the sort of graph that's suitable for this page. If someone is comfortable using R and has the time to recreate Mélencron's LOESS graph then I think that would suit the article best, if there's no longer consensus for the graph that's updated inline. Ralbegen (talk) 10:52, 22 April 2020 (UTC)

I wouldn't support a graph that cannot be edited fairly easily by others. Those graphs become a mess when Melencron left, and no-one knew how to fix them. --Comment by Selfie City (talk about my contributions) 13:19, 22 April 2020 (UTC)

If people are file-sharing anyway, then updating a CSV and running R would be as easy as doing that for an Excel spreadsheet. It's just about getting it set up again as we don't have Melencron's files. But Excel/Numbers/Sheets aren't really cut out for this kind of graph. Ralbegen (talk) 14:22, 22 April 2020 (UTC)

Hi. A few points. The graph shown isn't the LOESS graph, it is a SMA chart. It has the problem of 'jags': multiple average values on the same date. It has another problem which is that the averages are not centred, i.e they lag the data. Centering the averages solves both problems, using the survey date also helps ( fewer polls with the same date ). The LOESS averages were used in the run-up to the last election, and were ditched because they produced completely counter-intuitive results: the element of projection seemed to give current status with, e.g. parties in obviously the wrong order.

I don't really understand the problem with Excel: its charting facilities are excellent and it is commonly in use. I'd suggest that far more people are familiar with Excel than with R. Can you explain your concerns? If you look at the example instructions and link above, you'll see that the process really couldn't be much simpler. RERTwiki (talk) 09:08, 25 April 2020 (UTC)

Can we please go back to the moving averages chart??? I personally find it much more useful since it can show you exactly where changes in public opinion start, and because all of the previous graphs for other UK opinion polling pages are moving averages. Look, even if you want to keep LOESS, at least add the moving averages as well? It can’t do any harm to have both, surely? маsтегрнатаLк 10:21, 9 June 2020 (UTC)

Missing Polls

This one from Survation includes voting intention (fieldwork 31 January to 3 February). 2A01:E34:EE61:5F0:6934:E2AC:3D76:A74A (talk) 07:08, 12 February 2020 (UTC)

Exclusion of Bloomberg from polling averages

I noticed this reversion. I don't see why Bloomberg should be excluded. Shouldn't we keep all polls done by important pollsters for the sake of accuracy and up-to-date polling information? --Comment by Selfie City (talk about my contributions) 16:34, 28 March 2020 (UTC)

For the sake of reliable accuracy, it has in the past been agreed that only polling organisations who are registered members of the British Polling Council (BPC) are acceptable for this article. ~ BOD ~ 16:43, 28 March 2020 (UTC)

@Bodney: I can understand that. However, I think we can agree that Bloomberg is a reliable source and should therefore be included on our list of acceptable polling organizations. However, it's something upon which a consensus needs to be established one way or the other. --Comment by Selfie City (talk about my contributions) 13:37, 29 March 2020 (UTC)

We've included non BPC polls in the past, and for the sake of completeness I feel it's better to have too many than too little. Although I approve of the current policy, I am open to a compromise of a citation for non BPC polls if it puts the issue to bed. BrexitZZZ (talk) 13:59, 29 March 2020 (UTC)

I accept it is not as firm a rule as I thought, mostly the rule has been about stopping additions from non RS. In contrast, I am all for allowing exceptions especially when it is origin is indeed a reliable source. Apology for its removal. ~ BOD ~ 14:14, 29 March 2020 (UTC)

Are NumberCruncher not a BPC member? We have included their polls in the past. Certainly in the run up to the 2019 election.Cutler (talk) 18:26, 29 March 2020 (UTC)

Adding UKIP back to the main table?

We've just had the fifth poll in a row giving UKIP 1% of the vote. Surely it's worth considering putting them into the main table now? The "other" section being littered with citations isn't a good look. BrexitZZZ (talk) 12:20, 19 April 2020 (UTC)

I guess so, but is UKIP even worth a note at this point? Receiving 1% in a poll isn't noteworthy. --Comment by Selfie City (talk about my contributions) 16:05, 19 April 2020 (UTC)

Don't get me wrong, UKIP is a tiny party with a very optimistic ceiling of some 2%, but if pollsters are prompting it and get results for it, I feel it makes little sense to exclude/marginalise the results. The data is there, at a similar magnitude to Plaid Cymru. Citations are useful, but I feel should be limited to fringe cases like the odd BNP/WEP result that YouGov has historically shown. I'd like to see some more thoughts on this. BrexitZZZ (talk) 01:28, 20 April 2020 (UTC)

Yes. I've never heard of "WEP" (or many other British parties, not being British, and I’m not very familiar with any of the other minor ones mentioned here) and I'll have to do some research there, but that makes sense. However, I also think that given the fact the the UK has already left the EU, why include parties whose goals have already essentially been achieved? But I am glad to hear other opinions as well. --Comment by Selfie City (talk about my contributions) 13:37, 21 April 2020 (UTC)

I think that the parties reported in the headline results a polling company publishes are the most useful guideline for inclusion, and UKIP aren't amongst those for the ones I've checked. If they return to the spotlight and their results are consistently reported along with the larger parties then they can be retrospectively included for older polls. Ralbegen (talk) 13:47, 21 April 2020 (UTC)

I don't think UKIP are currently getting the sort of numbers that would warrant adding them to the table. Plaid Cymru are included not because of their vote share but because of their seats in Parliament. --Wavehunter (talk) 19:48, 21 April 2020 (UTC)

Right, I agree with Ralbegen here (and Wavehunter). --Comment by Selfie City (talk about my contributions) 13:21, 22 April 2020 (UTC)

I've been thinking, and I don't think any far-right or far-left parties should be included in the chart (including UKIP?). Extremist or near-extremist parties are unlikely to ever receive enough support to be worth including in the polls, so a permanent exclusion of those parties would set forward a sort of policy that would prevent future inclusion and debate. It's just a proposal, but otherwise we risk these extreme parties finding their way into the polling chart. --Comment by Selfie City (talk about my contributions) 16:24, 22 April 2020 (UTC)

It's a reasonable idea from SelfieCity, but we would then have to define far-right, far-left, extremist and near-extremist. A can of worms! I don't think UKIP would describe themselves as any of these things.--Wavehunter (talk)17:04, 22 April 2020 (UTC)

I do not think its our job to decide which political parties should be included based on how extreme their politics are (that is indeed a can of worms in itself), but leave it to the polls. Our task here is simply to record the polls. But I agree with Wavehunter, Ralbegen and Selfie City that currently they dont warrant inclusion. ~ BOD ~ 17:34, 22 April 2020 (UTC)

Political parties usually have their political position in the infobox on the right, often based on reliable sources. We could use the description in the infobox to decide each, in which case it wouldn't be difficult. Bodney has a point, though. --Comment by Selfie City (talk about my contributions) 17:56, 22 April 2020 (UTC)

Proposed replacement chart

Hi. With a view to trying to reach consensus on a better chart, I thought I'd open a new section here to give the discussion a little more visibility.

The chart is based on this Excel file, available in Dropbox Chart Excel File so that anyone can download it and update the chart. Instructions for update are on the first tab in the file.

The intention is to use Excel to create a PDF file which can be used as an image in Misplaced Pages. I've just discovered that I seem to need a new file name each time, so the instructions need to be updated to reflect that.

This is the current chart:

I will keep this chart up-to-date if we go with this, but clearly anyone else will be able to do so if I die or quit. Please let me know if you are happy to go ahead on this basis. RERTwiki (talk) 09:12, 28 April 2020 (UTC)

Hi. I've updated the chart to the new proposal. It's now about a week since the above with no further comment, I thought this would at least create some discussion. I've also updated the chart to use the same colour scheme as the previous chart, an made a minor update to the instructions. The excel file has thus changed, and the new file is available via this link RERTwiki (talk) 09:22, 5 May 2020 (UTC)

Thanks for doing that, having a Dropbox file is a good suggestion. I think the average should show the average of the previous ten polls, so the date should be at the end of those ten polls, rather than being in the middle of them. Otherwise we would have averages affected by polls in the future, which doesn't make sense. It also looks a bit weird currently having the trendline only for the middle points. Finally, you can upload a new version of an image by going to its page on Wikimedia Commons and going down until you see the "Upload a new version of this file" button (at the bottom of the "File history" section). Bellowhead678 (talk) 17:55, 5 May 2020 (UTC)

Some other formatting comments: I think it's obvious what the trendline is, so we can remove the AvCon, AvSNP etc. from the legend. We should also make the labels bigger to make them easier to read and get rid of the title (which is already covered in the description). We should also move the start and end dates so there is less white space. Bellowhead678 (talk) 17:58, 5 May 2020 (UTC)

Hi. Just came back to update for the weekend polls, and saw your comments. Thanks very much for the tip on updating files, I'll look into that for next week. There should be no 'editing the past' with the current positioning of the average. The last 10 polls actually represent the data centred on a date in the middle of the period they cover. That date will move forward as new polls are added, never update the past. I know it is conventional to plot the moving average at the endpoint, but it is wrong. It leads to the trendline not fitting the apparent curve of the individual polls - it would lag.

No problem making the labels bigger and reducing the white space on the time axis. I will trim the title, but would like to keep at least the date. I'm not sure about thinning the legend - those OCD as me might get twitchy. Thanks again, and next chart update will respond to these comments. RERTwiki (talk) 17:26, 11 May 2020 (UTC)

Hi. Updated the chart and consequently file, updated link is below in new section.

Took your advice ref file update. Fiddled with the title but didn't feel comfortable with no title or much less in it. Axis fonts are larger, average legend entries gone, less white space on the left. RHS white space is automatic, and needs to stay or else people will need to adjust the formatting for every update. Cheers,RERTwiki (talk) 10:09, 18 May 2020 (UTC)

To me the graph appears to have low quality (I'm sure there's a better name, but I've forgotten it). Should it perhaps be saved as a .svg file so that it looks clearer? --Comment by Selfie City (talk about my contributions) 13:32, 18 May 2020 (UTC)

I've together a graph that's more like Melencron's LOESS version, if there's any interest in that. I prefer its appearance: unlike Excel charts, we can have lines covering the entire polled period. The x-axis for a time series is clearer to follow with breaks at the start of months, rather than at days Excel chooses. The LOESS curves are smooth and seem sensible to me, though it's not difficult to change to another curve if they produce counterintuitive results at some point in the future.

Would there be any support for using this graph format (or something like it) on the page? Ralbegen (talk) 17:55, 18 May 2020 (UTC)

Could it be edited as easily as the current graph? --Comment by Selfie City (talk about my contributions) 18:40, 20 May 2020 (UTC)

Absolutely: I can put the R code and .csv as subpages on here or my userpage, so there's no need to go off-wiki. You can run the code online or download R for free, and users don't need to understand the code to run it, they just update the numbers and dates in the .csv and follow simple instructions. Ralbegen (talk) 18:55, 20 May 2020 (UTC)

Great! Then this seems like an improvement due to the quality of the image. --Comment by Selfie City (talk about my contributions) 21:04, 20 May 2020 (UTC)

I agree that this image is an improvement - would just suggest increasing the font size of the axis labels (so that mobile users don't have to zoom in/squint to read them) and changing the background from grey to white. Having a shared page with code that can be used by any editor is an excellent suggestion. Bellowhead678 (talk) 22:40, 23 May 2020 (UTC)

Would it also be possible to use this chart for other elections and polls?--BSMIsEditing (talk) 13:10, 23 May 2020 (UTC)

Thanks for the feedback! @Bellowhead678: I definitely agree that the text is too small relative to the rest of the chart. At your prompt I've tried changing the background colour to white, but that seems to make the breaks more obtrusive. I personally prefer the grey background with white breaks but I'm certainly open to further discussion on that. @BSMIsEditing: Yes, this kind of chart can be generally applied. There are already graphs that use the same kind of smoothing on pages like 2021 Senedd election. When I finalise my code and share guidance on updating it I can try to including some material on how to apply it more generally to put together similar charts for other elections, if that's something that'd be helpful?

On the basis of this discussion, I feel confident that there's some consensus to include this type of graph on the page. I'll implement an updated version of it, and then post my code with some explanatory notes on a subpage of my userpage at some point later this week. (There are still some changes I'd like to make to the code to make it easier for others to update). Happy to hear any feedback or dissenting views! Ralbegen (talk) 21:56, 24 May 2020 (UTC)

The formatting is a matter of taste, but I think the new chart looks good, and the addition of the election result is welcome.

Two points though - we know the LOESS smoothing has been misleading in the past in similar circumstances, and is therefore likely to be again in future. Consequently I would urge the use of a centred moving average for the average line now, rather than scramble to replace the LOESS curve when it goes pear-shaped. Among other things, the LOESS line has the 'edit the past' property that new data points are likely to move today's averages around, which is distasteful. A centred moving average makes the lag in the averages clear. I also believe that moving average weighted by sample size is a better average.

Could we have a link to the explanatory notes put in here? RERTwiki (talk) 08:51, 25 May 2020 (UTC)

Hi @RERTwiki:, thanks for the feedback. I'm really glad you like the format of the chart.

LOESS lines are used on a number other graphs of opinion results on articles on Misplaced Pages, and if you recall, when LOESS curves were discussed before, Melelncron clarified that the change in the curve was because of a parameter change rather than the addition of a new poll. The addition of a new poll has a minimal impact on the past curve, and as a majority of editors (including me) in that discussion pointed out, it's more useful to know where polling was based on all the information we know now rather than based only on the information we knew at each particular point. The issue of LOESS curves listing parties in a different order than reflected by polls at large is a possibility, but I think it's overstated. When LOESS was implemented last time it was around the time of the 2019 European Parliament election, when UK general election polls were behaving in exactly the way that would cause that issue (with the relevant discussion here. In any case, I believe it's something that would have been resolved once later polls were included.

By comparison, simple moving averages create very noisy curves that (in my view) obfuscate rather than illuminate the actual trends in opinion polling. I am of the view that LOESS curves remain the most suitable option. (There's a version of LOESS called compositional LOESS that's designed for opinion polls in particular presented here but that paper's beyond me. If anyone can implement that in R then that method would be my preference.)

Finally, yes, I will definitely share a link to my notes here once I've finalised them. There are a few more things I'd like to sort out in the code before I do that though, in order to make it as straightforward as possible for other users to update it. Ralbegen (talk) 11:34, 25 May 2020 (UTC)

Thanks for that, might make sense to replace the section below with the link to the previous chart with your info.

Yes, the 2019 polling was the occasion where LOESS broke down which I was referring to. These data are politically sensitive, and it is not sufficient to say that future polling would (might have) corrected the situation. People will (did) immediately become annoyed that things which they cared a lot about were being mis-represented. I agree this was an extreme circumstance, but I really don't think it is worth the risk to include a LOESS line up to the current date which is almost certain to be corrected, albeit usually only slightly.

One reason the moving averages have given noisy curves in the past is the way they have been constructed. Plotting averages against publication date creates situations where there are often multiple values for a given day. This creates 'jags'. If instead you plot against the weighted average sample date-time, there are very rarely multiple values for a given time, and the data is much smoother. Thus I think the sample-weighted moving average of polls plotted against the sample weighted moving average of sample dates produces smooth (enough) curves, correctly represents the timing of the average, gives a value which is fixed for all time, and is very simple for the reader to understand. I think it's better for those reasons. I still have misgivings concerning the possible double-entry of data into a CSV file and the table in the article, which is an obvious source of potential errors, but I guess your instructions will shed more light on that.

All that said if you are intending to keep the chart up to date, it's certainly a big improvement on where we were a few weeks ago, and I'd be pleased to see you do that.RERTwiki (talk) 11:15, 26 May 2020 (UTC)

Current link to Excel file for chart

This is a link to the current version of the excel file used to create the latest version of the polling chart. The intention is to update the file and link here whenever the file changes (not for each new chart). RERTwiki (talk) 10:09, 18 May 2020 (UTC)

Redfield & Wilton Strategies subnational polling

I have noticed that on the R&WS website that there is a section for different regions of the UK, which may potentially be a source for subnational polling. But Wales will most likely be excluded as there seems to be an error as it displays the SNP there. — Preceding unsigned comment added by BSMIsEditing (talk • contribs) 18:42, 30 April 2020 (UTC)

BSMIsEditing - The sample size for the regions in national polling is generally too small to be of any use (often below 100 for the smaller regions), and often isn't weighted by age, gender, social class etc, as the national sample is. PinkPanda272 (talk/contribs) 07:00, 8 June 2020 (UTC)

Proposed Merger

The following discussion is closed. Please do not modify it. Subsequent comments should be made in a new section. A summary of the conclusions reached follows.

The result of this discussion was Don't Merge. PinkPanda272 (talk/contribs) 11:14, 12 June 2020 (UTC)

I am proposing to merge Leadership approval opinion polling for the next United Kingdom general election into Opinion polling for the next United Kingdom general election. We are at least 4 years from the next election, and I don't think it is necessary to have two pages with similar content, and they can easily be split back out if need be nearer the time. As context, there are two opinion polling pages for the 2015 and 2019 elections, but only one for the 2017 election. Pinging @The Scandinavian Chestnut, Sofia9, Ralbegen, Mattftom, JackintheBox, and Impru20: Thanks, PinkPanda272 (talk/contribs) 07:00, 8 June 2020 (UTC)

I do not feel strongly either way on this question, but I would have thought both articles are only going to grow (and indeed Leadership approval opinion polling for the next United Kingdom general election is already quite big), so let's keep them separate. Bondegezou (talk) 08:56, 8 June 2020 (UTC)

Ugh no, you in the UK have already managed to keep both articles separated in a nice way and merging the two of them will only result in a single but oversized and impractical article. Also consider that both of them will only grow bigger (think that they already comprise 30kB and 40kB respectively and it's just slightly over six months since the last election. The parliamentary term can last up to five years. It's very likely each of them will reach or surpass the 100kB mark with the data of this year alone). Also, content is not "similar"; one includes electoral opinion polling whereas the other focuses on leadership opinion polling. Would be like merging the 2019 general election article with the 2019 European Parliament election article just because these two share being elections held throughout 2019. Impru20 09:24, 8 June 2020 (UTC)

I don't think any of the WP:MERGEREASONs apply enough here to justify merging the pages, especially as Bondegezou and Impru20 point out, both are likely to grow. The 2017-19 articles are both enormous for a two-year term, and the 2010-15 article was so big it had to be split into five articles, including the leadership approval article, totalling more than a million bytes. One of those articles is still the 34th biggest page in the encyclopedia. I think it'd be wisest to keep them apart for now. Ralbegen (talk) 10:53, 8 June 2020 (UTC)

Thanks for the responses. I will keep this discussion open for a further day or so in case others wish to contribute, but I think we have a relatively strong consensus against merging. At present there is not much linking the two articles, apart a mention in the 'see also' section. Would it make sense to add the 'leadership approval' article to the infobox, to read (for example):

It may also be an good idea for the infobox to be standardised across the recent election/opinion poll articles, as I have noticed that some pages (such as 2015) don't have a link to the next (2024) election, while this page doesn't link back as far as 2010. Thoughts? PinkPanda272 (talk/contribs) 11:50, 8 June 2020 (UTC)

That's an interesting idea. The current infoboxes for opinion polling articles aren't really infoboxes so much as navigation aides and I wonder if something more like Template:US 2020 presidential elections series could be the solution here? Ralbegen (talk) 12:03, 8 June 2020 (UTC)

Use of Bristol Elects constituency polls

They don't seem especially less professional than other university pollsters which wikipedia cites for US articles. BPC membership is apparently the usual standard for inclusion here, but is that really necessary if we know the source is reliable and not deemed to be faking the data by other reliable sources? It's not the usual wikipedia standard for inclusion of sources.

We could instead include a key indicating whether a pollster had BPC membership or not. I think these polls add value in the absence of further evidence suggesting they *are* faking their data. - PutItOnAMap (talk) 17:51, 14 June 2020 (UTC)

Since when is there a consensus for requiring BPC membership? I remember this requirement being rejected back in 2017 (see Talk:Opinion polling for the 2017 United Kingdom general election/Archive 4). Rami R 17:08, 14 June 2020 (UTC)

• Unassessed Elections and Referendums articles
• WikiProject Elections and Referendums articles
• Unassessed Politics of the United Kingdom articles
• Mid-importance Politics of the United Kingdom articles