Talk:Opinion polling for the 2011 New Zealand general election

Latest comment: 12 years ago by Trevva in topic Last minute request

Graphs

edit

Great to see that you guys are into it already. I'll modify my graph script sometime in the next week or two, and then we're back into it again! Also, one suggestion - would it be an idea to add the most recent poll to the top of the table, so that this is the first thing you see? Or keep it so that it's at the bottom? Looking forward to working with you all again! --Trevva (talk) 15:07, 30 June 2009 (UTC)Reply

Yes, I think that the more recent results should appear at the top....Adabow (talk) 04:41, 1 July 2009 (UTC)Reply
Opposed. Why break with tradition?KiwiDave (talk) 06:35, 5 August 2009 (UTC)Reply
Because it is easier for readers to get the information–one can just access the page and at tyhe top of the page is the most recent opinion polling data. Adabow (talk) 08:30, 5 August 2009 (UTC)Reply

So, it ended up being a bit more than a few days. But I finally managed to put a new version of the graph up there. I'm very open to suggestions for improvements etc if there's something else you'd like to see on it --Trevva (talk) 08:45, 23 November 2009 (UTC)Reply

Unfortunately there was a major change made to the date formatting style within the last month or so, so that "16 January - 21 January 2010" has been replaced with "16-21 January 2010". I agree that the result is more human readable, but it makes it less machine readable unfortunately, and has broken the R script. I'll see if I can make some mods in the next few days and get it working properly again --Trevva (talk) 07:33, 18 October 2010 (UTC)Reply

Fixing those date style changes was tricky - I had to completely reconsider the date extraction algorithm, but what I have now works a lot better. Fingers crossed for no more changes in style! --Trevva (talk) 15:13, 2 February 2011 (UTC)Reply

Hi there, first off, sorry total noob with discussion etiquette so sorry if this is not the appropriate place or not formatted correctly. What I have is a request. Under the main graph could we generate a graph with a much shorter timeline? Now we're close to the election I'm really interested to see week by week changes in polling. Perhaps 3 months before polling day would be a good start point. Obviously a graph with a shorter timeline will be less accurate and more prone to fluctuation, but with the current one I really struggle to see how the parties are doing with their campaigns. Thanks. Jonathan. — Preceding unsigned comment added by 121.98.108.49 (talk) 22:42, 13 November 2011 (UTC)Reply

POV

edit

"9 May 2011 – The "Zero Budget" is released, with $1.2b cuts to spending to help fund the rebuilding of Christchurch after the 22 February earthquake" The Chch earthquake only contributed to about 10% of the government's deficit I think this is pretty POV — Preceding unsigned comment added by 203.79.117.138 (talk) 09:49, 6 June 2011 (UTC)Reply

I agree. I've replaced the "to help fund the rebuilding of Christchurch after the 22 February earthquake" part with "scheduled over the next four years". All of the comments within the table should ideally cite a source too. --Avenue (talk) 12:46, 6 June 2011 (UTC)Reply

Horizon comparable?

edit

Is the Horizon poll comparable to the others? It seems it asks a very different question (asking for preference in undecideds). More worryingly, it seems from the webpage that HorizonPoll has self-selection bias, and so is a poll of people interested in doing online polls for a chance to win an iPad, not a representative poll of all New Zealanders. It does some fiddly weighting for past preference, true, but is still a fundamentally different sample space. There are enough results from the (comparable) other polls, that the Horizon poll is a confusing factor in this list of polls; it does not track with other polls. — Preceding unsigned comment added by 130.216.148.210 (talk) 00:36, 25 July 2011 (UTC)Reply

I agree, this poll looks odd. Polls on either side are consistant. Suggest it is treated with care, pending further investigation as to its methodology and philosophy.

They state:

METHODOLOGY NOTE:

Horizon party vote polling differs from other polls which telephone voters on landlines at random, and express decided voters as a percentage of 100 and do not weight results on 2008 party vote.

Horizon selects its respondents by e-mail to match the New Zealand population at the 2006 census. Respondent samples are further weighted by age, gender, ethnicity, personal income, region and party vote 2008 to provide a representative population sample aged 18+.

Party vote respondents are further filtered to exclude those not registered to vote and not intending to vote. Horizon also polls larger numbers of people than most other polls, in order to better indicate support for smaller parties, which have been vital in forming coalition governments under MMP.

What is interesting on its site is a table showing movement between what people voted for in 2008 with their preferences now. NealeFamily (talk) 20:35, 26 July 2011 (UTC)Reply

I agree, this poll is unreliable. Anyone can sign up online to be a part of the poll and they offer gifts to those that join. Even if it was reliable it distorts the graph because of its different methodology. It should be removedSobitemybum (talk) 07:56, 1 August 2011 (UTC)Reply

I would also agree with the consensus - if its a self-selecting poll, it probably shouldn't be in there. The fact that they get a result that is many stand deviations away from the others also suggests they're doing something "different" (bad different, not good different ;-) Shall we skip over it? --Trevva (talk) 09:48, 3 August 2011 (UTC)Reply

I wouldn't say that the Horizon polls is necessarily "bad different", but it is clearly incomparable, and so it needs to be analysed separately. Yes, Trevva - please skip over it.203.173.244.240 (talk) —Preceding undated comment added 09:06, 6 August 2011 (UTC).Reply

Updated the figure to reflect the removed Horizon poll. --Trevva (talk) 07:09, 8 August 2011 (UTC)Reply

Err I reverted the deletion of the horizon poll as it was done by an IP address with no explanation in the Edit summary. I see here that there is some discussion about this, so I'll revert it back again. Onco p53 (talk) 02:15, 9 August 2011 (UTC)Reply

Party polling graph

edit

The polling average for National is superimposed on the graph legend. It needs fixing to make it easier to read. -- Alan Liefting (talk) - 21:32, 26 July 2011 (UTC)Reply

I have dipped my toe in to the R script and moved the legend to between the Greens and Labour. I hope its not too clunky. Mrfebruary (talk) 07:34, 29 July 2011 (UTC)Reply

Sorry, I knew this was going to be an issue at some point, but forgot about it. I have put a modified version of the script up on wikimedia so that it places the legend at the very bottom - hopefully this looks a little tidier. If you'd like to see some other changes/improvements, please just let me know... --Trevva (talk) 09:52, 3 August 2011 (UTC)Reply

No problem. Compliments on your script. It worked flawlessly in downloading and cleaning the data.Mrfebruary (talk) 07:36, 4 August 2011 (UTC)Reply

NZ First over 5% threshold

edit

The caption of the graph says that parties clearing the 5% threshold get included in the graph. NZ First, although polling low most of the time, has cleared the threshold twice. We need to either include NZ in the graph, or change the stated criteria for inclusion. Since the same criteria are used in the pages for the 2005 and 2008 elections, I think it would be best to include NZ First here. Ridcully Jack (talk) 23:12, 10 August 2011 (UTC)Reply

I've changed the graph caption to say "consistently polled above the 5% threshold". However, I have no objection if anyone regenerates the graph to include NZ First and reverts my change.-gadfium 02:31, 11 August 2011 (UTC)Reply
I think that is a good move. Having NZ First would clutter the graph (as also stated in the caption). I must declare a conflict of interest here as a supporter of the Green Party. -- Alan Liefting (talk) - 04:06, 11 August 2011 (UTC)Reply
I agree. We can revisit this if NZ First poll over 5% a few more times. Ridcully Jack (talk) 08:30, 11 August 2011 (UTC)Reply
... although, having generated the graph with NZ First included, it is not cluttered as the above poster suggests; no more so than the 2008 graph was; the gap between Greens and NZ First is clear. Ridcully Jack (talk) 09:35, 13 August 2011 (UTC)Reply
Why don't you upload the graph which includes NZ First (with a different file name) so we can compare them?-gadfium 20:43, 13 August 2011 (UTC)Reply
 
Here's what the graph looks like with NZ First included. Ridcully Jack (talk) 00:37, 14 August 2011 (UTC)Reply
Looks a little crowded with both parties. NZF have only peaked over 5% in the occasional poll - may need the party to peak over 4% minimum over a matter of polls for inclusion.
Maybe a separate graph for parties polling regularly under the 5% threshold, with New Zealand First, ACT, Maori, Progressive, United Future and Mana, as a kind of magnification of the bottom 5%. Would show trends in minor parties better than reading them from a table. Lcmortensen (mailbox) 01:26, 14 August 2011 (UTC)Reply

I totally agree with Lcmortensen above. It would be great to see a seperate graph of the minor parties, expecially with the amount of interest they bring to the election. ----

I would commend Ridcully Jack on taking the initiative to add the graph of the smaller parties. However, having it as a thumbnail doesn't work out so well unfortunately - it tends to disappear behind the larger graph on my 1024x768 display. I see two alternatives: 1. make it a full size graph, on the same size as the other graph. 2. Combine the two figures into one, alternative layout. Here's an example of 2 that I test-drove around 2009... Alternative split graph design Variations on this theme are also possible. My vote goes for option 1. --Trevva (talk) 12:58, 21 August 2011 (UTC)Reply

I also prefer option 1. Ridcully Jack (talk) 23:35, 22 August 2011 (UTC)Reply

New "minor parties" graph: as suggested

edit

I've added a new graph showing a comparison of the minor parties, following the above suggestion. The criteria I've used for inclusion is polling over 1% (but not consistently over 5%), which means a party has a chance of bringing a list MP into parliament if it should win an electorate seat. As it stands, JA's Progressives and Mana would not bring any list MPs into parliament should they win an electorate seat. (It's not clear that the Progressive Party is even contesting the 2011 election). In fact, parties need to get to (around) 1.5% to bring in a list MP on the strength of also winning an electorate, so setting the cutoff at 1% is probably generous, and it could be a little higher. Ridcully Jack (talk) 02:39, 14 August 2011 (UTC)Reply

Hi Ridcully Jack. I'll have a look at the script and see what I can do to get Mana party in. Parsing the HTML is a bit of a nightmare, so its possible that there is something odd in there! --Trevva (talk) 10:51, 20 August 2011 (UTC)Reply

Mana is also working in the script now. I was a bit in doubt what to do with their polling before party creation, but decided to set it to NA (missing) rather than zero. New version of the script (see graph description) has the option to produce either a small parties graph, or a major parties graph --Trevva (talk) 19:32, 22 August 2011 (UTC)Reply

Just to follow up, I have included four parties in the small parties graph - ACT, NZ First, Mana and Maori. I was probably too generous including Mana, but anyway. United Future hasn't made it above 1.0%, so they're out. --Trevva (talk) 19:43, 22 August 2011 (UTC)Reply

I've tweaked it a little, to see another decimal place. Also, I've put United Future in, as I realised that parties who are reasonably likely to win an electorate (Dunne, Harawira) who poll below 0.4% will be in overhang, which is also interesting when following small parties in the polls. Ridcully Jack (talk) 23:15, 22 August 2011 (UTC)Reply

meta colours in R?

edit

Is there any way the R code can be modified to use the meta colours for political party colours? Some of the colours are a bit off (especially the ones I've used); it would be nice to be able to use #F5E4BC style colour names. Ridcully Jack (talk) 08:18, 21 August 2011 (UTC)Reply

Yes there is! R accepts hex colours, so I can just read them directly from the table... I'll look into it. --Trevva (talk) 12:15, 21 August 2011 (UTC)Reply

Done. The new version of the script (see the image description) uses the same meta colours as the table, and also takes the party names from there as well. --Trevva (talk) 19:30, 22 August 2011 (UTC)Reply

Noteworthiness of table 'news item'

edit

This row appears in the poll results:

31 July 2011 – New Zealand First announces at annual convention that it will repeal the anti-smacking legislation if elected to Parliament

Is this really noteworthy? Lots of political parties release policies, especially nearing elections. Either every policy announcement is notable (and you won't be able to find the poll data for the policy announcements) or policy announcements need to meet a relatively high standard of news-worthiness. Tax packages, budgets, by-elections, changes in leadership, expulsions, and controversies that might affect voter perceptions are valid inclusions. This seems not to be. Ridcully Jack (talk) 10:03, 22 August 2011 (UTC)Reply

I agree, this is not of sufficient importance. These lines should be reserved for events which make a lasting change to the political landscape.
Is the line "9 June 2010 – Ministerial credit card statements were made public, revealing misspending by some Labour MPs." suitably neutral? I thought MPs of many parties were shown to have exercised doubtful judgement on use of credit cards.-gadfium 22:04, 22 August 2011 (UTC)Reply
The credit card statements going public isn't all that noteworthy - I can't find anywhere I can cite from within wikipedia to make it noteworthy. It was ministerial spending that was questionable, so it might just have been Labour? There's no sign in the graph that Labour lost support due to the 'scandal'; in fact they look to be trending up in June '10, so in retrospect it didn't have much effect on the politcal landscape. I support removing it, or at least more neutral wording. It definitely needs a citation. Ridcully Jack (talk) 22:27, 22 August 2011 (UTC)Reply
Credit card statement removed. Lcmortensen (mailbox) 05:41, 23 August 2011 (UTC)Reply
By-elections, changes in leadership, expulsions - anything that changes the composition of parliament or its parties are noteworthy. Do need to include the changes in Labour and Green leadership since the election (Phil Goff replacing Helen Clark, and Metiria Turei replacing Jeanette Fitzsimmons), which I'll add shortly. Lcmortensen (mailbox) 05:56, 23 August 2011 (UTC)Reply
I wonder if there is a date for when it became clear that JA's Progressives would not be contesting the 2011 election?
Also, there have been several changes of List MPs due to retirements etc; for instance are David Clendon (2/11/09) and Gareth Hughes (16/10/2010) were sworn in as MPs as replacements for Sue Bradford and Jeanette Fitzsimoons respectively. These aren't particularly noteworthy, but do change Parliament. There are several other list MP changes; these were just the first two I came across. Ridcully Jack (talk) 06:09, 23 August 2011 (UTC)Reply
Maybe leave the List MPs off, unless of course they are leaders or deputy leaders (or former leaders) of their party. Changes in list MPs are almost routine, unlike electorate MPs which require the big fuss of a by-election. Lcmortensen (mailbox) 06:52, 23 August 2011 (UTC)Reply
I've removed the NZ First announcement. Anyone who wants it back might first explain why it is sufficiently important. Ridcully Jack (talk) 03:07, 25 August 2011 (UTC)Reply

News item inclusions in table

edit

It might be a good idea to have a clear list of what is sufficiently noteworthy to be included in the poll results table. Since the table is primarily a record of polling, and not a record of politcal events, events which change the politcal landscape significantly should be included. At the moment, the table includes:

  • Changes of party (co-)leadership (for polling parties)
  • Constitutional events (swearing in a Prime Minister)
  • By-election winners
  • any MP leaving a party (becoming an independent)

Also, we currently have some events which are less about the makeup of parliament and more about significant party platforms

  • National: partial state asset sales
  • National: 2011 zero budget
  • Labour: race-related "nationhood" speech (is this item NPOV? it reads like a criticism)
  • Labour: tax package including capital gains tax

It seems that deciding which policies get included is less clear-cut, and might need discussion. Ridcully Jack (talk) 03:28, 25 August 2011 (UTC)Reply

Some ideas for events
* The 2009 and 2010 budgets - all budgets are significant for polling, as they can make or break a government
* The 2009 smacking referendum returning a landslide "No" result and the abstained response of Parliament.
* Significant and/or controversial legislation
* The 4 September 2010 and the 22 February 2011 Christchurch earthquakes - in relation to how the Government's response affected polling (the Roy Morgan poll immediately following the 22 Feb quake put Government Confidence at its lowest since the government was elected)
And yes, criteria is not exactly clear-cut - its a matter of what's notable and what's not. Lcmortensen (mailbox) 13:06, 3 September 2011 (UTC)Reply
Oh and for 26 October 2011 (not quite yet), we should note Writ Day and the official start of the campaigning season - polls after Writ Day could swing wildly with election promises. Lcmortensen (mailbox) 13:38, 3 September 2011 (UTC)Reply
I've added lines in the table for both earthquakes and the referendum. I'm not convinced that the 2009 and 2010 budgets are particularly noteworthy. Ridcully Jack (talk) 22:37, 6 September 2011 (UTC)Reply
Can I suggest also the inclusion of the Rena oil spill off the coast of Mt Maunganui? I believe that it's an issue that will be reflected in the poll results, for example, as high as 11% of National supporters said that the incident might affect their opinion. — Preceding unsigned comment added by 101.98.143.173 (talk) 04:22, 21 October 2011 (UTC)Reply
Re Rena, see below. Schwede66 04:29, 21 October 2011 (UTC)Reply

Fairfax polls

edit

Hopefully no one minds the inclusion of the Fairfax Media Research International polls. I have checked the criteria used to conduct the poll, and it is nearly identical to the four other polls - telephone poll of 1000 voters, and the question looks very similar to the one TV3 uses.

A bit high on the National figures compared to the others (then again, Fairfax isn't exactly politically neutral), but the minor parties are comparable. Lcmortensen (mailbox) 13:06, 3 September 2011 (UTC)Reply

I think it's reasonable to include them. I can't find any historical data on Fairfax polls, but if there are polls from before July, they would give more data points for the graphs, which would be a help. Ridcully Jack (talk) 22:12, 5 September 2011 (UTC)Reply

MV Rena

edit

It seems that the 2011 Tauranga oil spill is noteworthy for the table.

Any suggstions for how to neutrally describe the event? (POV will easily rear it's head when discussing government response, but it is possible that perceptions of the speed of the response will show in the polls). Ridcully Jack (talk) 04:23, 19 October 2011 (UTC)Reply

It says in that article under the heading 'Political consequences' (modified by me to make it suitable for this article): "The disaster occurred only seven weeks before the 2011 general election (which will take place on 26 November 2011) and is expected to have an effect on the campaign.[1] On 14 October it was reported that the disaster had caused a 4% drop in the governing National Party's polling on the iPredict prediction market.[2]" That sounds neutral enough to me. Schwede66 04:46, 19 October 2011 (UTC)Reply
We do have to consider it is slightly confounded with the double credit downgrade by both Fitch and Standard & Poor's on 30 September, and the government's response. Both I think should be included. Lcmortensen (mailbox) 01:58, 21 October 2011 (UTC)Reply
  1. ^ Audrey Young (14 October 2011). "Rena oil-spill emergency winds up political leaders". The New Zealand Herald. Retrieved 19 October 2011.
  2. ^ "Rena disaster hurts National - site". Stuff. 14 October 2011. Retrieved 19 October 2011.
I don't think we should cite the shift on ipredict, as its numbers are different to polling numbers (again, like Horizon's, not wrong, just different). If there is a shift, we'll see it in the next run of polls. Ridcully Jack (talk) 20:50, 24 October 2011 (UTC)Reply

Poll dates

edit

Should the poll date be the date the poll is published or the date(s) that the poll was conducted? Since the difference is usually on a few days it probably isn't a huge problem. 115.188.243.92 (talk) 20:21, 1 November 2011 (UTC)Reply

The problem with aiming to use the dates polling was conducted is that some polling doesn't make this public. The convention used seems to be that if you can find the dates the poll was conducted, to list the poll with those dates, sorted in the table by the end date. (As an aside, I'm not sure how the R code generating the graphs uses date ranges, but I suspect it uses the start date.) Ridcully Jack (talk) 21:58, 1 November 2011 (UTC)Reply
Where a range exists, the R code takes the middle point (mean). --Trevva (talk) 20:02, 6 November 2011 (UTC)Reply

Conservative Party

edit

Some work may be needed to include another column for the Conservative Party. They poll at 1.1% in the latest Herald poll, significantly higher than some of the others in the table. Of course, this could be a one-off, but it will be interesting to see whether they also feature in tonight's TV One poll. If anyone with experience with wiki tables would like to add a column (if you think it should be added), I will try to get the R code sorted to include them in the minor parties graph. (Some fiddling similar to Mana's shortened trend line will be required, and Trevva, the originator of these lovely graphs, seems to have been away from Wikipedia since August). Ridcully Jack (talk) 03:13, 3 November 2011 (UTC)Reply

Back again - you guys are doing a great job of running the graphs, so I'm just sitting back and watching it all unfold! I think that the addition of an extra column shouldn't break anything, but let me know if you strike problems. --Trevva (talk) 20:05, 6 November 2011 (UTC)Reply
The Conservative Party column is back. We need consensus on adding columns to this table. I don't think that setting a particular number that at party must poll at is the right approach to take, but a party needs to be represented in a significant number of polls (or alternatively be represented in parliament). There have been 22 polls since August and 13 in the campaign period. I think it would be reasonable for party to appear in at least 8 polls (above zero) before appearing in the table. Ridcully Jack (talk) 07:55, 18 November 2011 (UTC)Reply
The Conservative Party has been polling higher than Mana in some polls. And higher than United Future in most. It has been represented in most polls for the last few weeks. I think it qualifies for a column. I don't see anyone else on here saying in previous discussions that the Conservatives shouldn't be on the graph. They aren't a one off so therefore I'm reverting it backUser:Luke96241 (talk) 08:16, 18 November 2011 (UTC)Reply
I take your point - but one could argue that with a foothold in parliament, Mana and United Future are more noteworthy. However, I am inclined to revise my threshold above - partly because it would be quite retrospective, and require more work to add extra data from polls back in later when a # of polls threshold was reached. It's clear that the Conservative Party is 'on the radar' to pollsters, appearing in 3 of the 5 polls, and regularly in NZHerald.
I think we should keep the Conservative column in the table. Ridcully Jack (talk) 08:21, 18 November 2011 (UTC)Reply
(But I can't add them into the minor parties polling graph, in case you were wondering. There just isn't data over a wide enough time period, and the Loess smoother function chokes. The five blue dots for Conservative get lost in the clutter.) 08:23, 18 November 2011 (UTC)

A Big Thanks

edit

Just a quick note of thanks to all the contributors keeping this page up to date. It's great to be able to watch the trending across such a range of polls over such a long time. Lisiate (talk) 02:33, 10 November 2011 (UTC)Reply

Big thanks seconded. an invaluable tool. User:Mistywindow — Preceding unsigned comment added by Mistywindow (talkcontribs) 23:11, 17 November 2011 (UTC)Reply

Legalise Cannabis Party

edit

I think it is fair to include a column for the Legalise Cannabis Party, especially if Progressives, Conservatives and Mana are included. Alcp are currently polling higher than UF or Mana and Progressives don't even exist anymore. Conservative is a one off, so under these criteria ALCP can be included otherwise it is a double standard. Anyone who doesn't think they should have a column on the table please justify it here coherently before removing the column. Otherwise a 3rd opinion will be sought. 118.90.92.253 (talk) 23:27, 16 November 2011 (UTC)Reply

I suggest that you need to gain consensus on the talk page before making fundamental changes. Don't implement fundamental changes and then ask others to justify why it shouldn't be done that way. Your approach won't get you very far on Wikipedia. You are a new editor so you won't be aware of WP:3RR, so here's a friendly suggestion to read that page before you do any reverts, as you have already violated it. Schwede66 23:57, 16 November 2011 (UTC)Reply
There are several problems here. United Future, Mana, and Progressive are existing parties in the 49th Parliament, so justify inclusion along with National, Labour, Green and ACT. New Zealand First is a consistent outside poller. Conservative is not exactly a one-off, as they are currently actively polling with more than one polling company and are polling above 1%. ACLP is only polling with one company and below 1%.
If we see them polling with another company and above 1%, then they would deserve inclusion. Lcmortensen (mailbox) 00:06, 17 November 2011 (UTC)Reply
Your two guys opinion is hardly consensus, so I don't think your arbitrary rules for inclusion that you have stated above can be taken as the end all and be all of the format of this page. I have read all the above discussion and I don't see any discussion of rules for inclusion in this table, or any indication that consensus has been reached surrounding this. So I don't see how you can cite the above rules. ALCP has contested every MMP election and at 0.8% account for one MP's worth of the vote. They obviously meet the NZ Herald criteria for inclusion as they have a dedicated column in that table, so in terms of notability Wikipedia guidelines would dictate that if the Herald thinks its notable then it's notable and your guys personal opinions have no place in it. PS I'm changing it back now118.90.92.253 (talk) 02:06, 17 November 2011 (UTC)Reply
Also I am aware of the 3RR, and you guys tag teaming me to get around it will not be looked upon favourably when a third opinion is sought.118.90.92.253 (talk) 02:07, 17 November 2011 (UTC)Reply
I have filed a 3RR report. Schwede66 03:24, 17 November 2011 (UTC)Reply
The 3RR report resulted in a 24 hour block for User:118.90.92.253. The admin who blocked the user advised that "if the user socks under a different IP, by all means request semi-protection." Schwede66 20:30, 17 November 2011 (UTC)Reply
I've seen "or 3O will be sought" threats before[1]. That is most likely a certain list candidate of the ALCP, and I've opened a SPI case over their abuse of multiple accounts. I also suspect they've received assistance from User:Alan Liefting, who contacted me off-wiki right after one edit war, and has made his views on cannabis clear. XLerate (talk) 08:26, 18 November 2011 (UTC)Reply

Last minute request

edit

Would it be possible to get a compressed graph showing the last month of the campaign? This might show up trends that get lost in the long term series. — Preceding unsigned comment added by 203.97.96.195 (talk) 05:52, 17 November 2011 (UTC)Reply

I'm not sure that we need more graphs for the page; I don't have time to modify the R code to do this, but others might. Ridcully Jack (talk) 08:09, 17 November 2011 (UTC)Reply
 
It is less the scale, and more the twitchiness of the smoothing line. This is only this year's data, with the loess set to 0.2 rather than 0.5, and I suspect it shows what you think that it should, that there is actually more marked change in the last few weeks. --Limegreen (talk) 09:52, 17 November 2011 (UTC)Reply
I also just tried not altering the axis, but just the smoother, to both 0.2 and 0.3. Interestingly, 0.3 produces quite a different interpretation than 0.2 (Labour appear to be losing nothing on 0.3), and the change for National and Greens seems less extreme. --Limegreen (talk) 09:58, 17 November 2011 (UTC)Reply
It might be worth considering making the smoother "twitchier" now that there are more data points for the main graph. I will generate graphs with 0.25 perhaps, later today (unless someone beats me to it). I'll put them here first. Ridcully Jack (talk) 20:36, 17 November 2011 (UTC)Reply
I've had some time to reflect on this, and I think the case the for making it somewhat twitchier is there. I also think on balance that too much twitchier may give too much weight to current polls, as a certain amount of inertia seems reasonable. Interestingly, this graph on the talk page has been picked up here [2] and here [3]. I broke my minor party script trying to put the Conservative Party on, before they were removed, so might be best if you do it. The difference between 0.2 and 0.25 is quite strong (0.25 more closely resemble 0.3) and I think 0.3 gives a still relatively smooth line, but I'm open to debate. -Limegreen (talk) 23:51, 17 November 2011 (UTC)Reply
I think I'll change the graphs over to "twitchier" versions later tonight; but I'd like to hear from others who view the graphs regularly or contribute to / follow this page to check there is some consensus here. Ridcully Jack (talk) 00:44, 18 November 2011 (UTC)Reply
I agree. My particular concern is that I'm not too familiar with the loess function, and have implemented the change be reducing the value of "span". My reading of the R helps suggests this isn't inappropriate. On an interpretation level, I'm OK with it twitchier (although it looks like hell on the smaller party graph, largely because that data is much noisier), as it seems to reflect an evident trend. -Limegreen (talk) 01:11, 18 November 2011 (UTC)Reply
From my understanding, there isn't really a good way to choose the "span" argument to a Loess function - the default is 2/3, which means that it uses 2/3 of the data to do the local-regression. When I first wrote the script, I just choose a number that looked good. There's no reason why you can't just choose another number - what you have there looks pretty good to me.

On a related note, I have implemented a version of the script using a GAM (Generalised Additive Model) spline smoother. This is theory should be superior to a loess, because it chooses its "twitchiness" (great word, by the way) automatically. The problem is that it still doesn't respond to the most recent points, which, it seems are being reflected in multiple sources. Win some, lose some, I guess! We could also use this approach to correct for biases between surveys, which it seems is an important issue. For the moment I've left it out though--Trevva (talk) 00:45, 21 November 2011 (UTC)Reply

Tea Tape scandal / Green stickers on billboards / Show Me The Money / ...

edit

What about also including: "Phil Goff fails to respond to John Key's taunt of "show me the money" with regard to uncosted election promises." "700 National billboards are defaced by a group organised by a Green Party member closely linked to Russel Norman." But actually I argue that including any events in main campaign period is likely to become a flurry of events and policies of lessening importance. NPOV will be difficult to achieve. Ridcully Jack (talk) 03:21, 18 November 2011 (UTC)Reply

That stuff can be put in the New Zealand general election, 2011 article. -- Alan Liefting (talk) - 04:17, 18 November 2011 (UTC)Reply
There is a section in the New Zealand general election, 2011 article entitled "Campaigning" which is currently empty, and it would be best to put most of these news items here. The Tea Tape scandal is an exception and should be included in the listings because of its notability and controversy (and the fact it went international). Lcmortensen (mailbox) 10:26, 18 November 2011 (UTC)Reply