renton

Not really, actually.

Be fair now, it's not like you were all over the thread with your in depth analysis when Survation and ICM showed huge Yes swings in January and PB showed a 2 point drop in NO in February - but as soon as Ipsos shows a number you like, you are back in here with your usual 'gonna get humped' rhetoric.

Well, actually it's about Scots re-asserting sovereignty over their nation, but thanks for the very vivid demonstration of how different wording can give different perceptions.

Pollsters publish their data tables to differing degrees.

And Ipsos are the only ones showing such a huge No lead (and smallest undecided) and yes, who commissions it makes a difference (because they influence what is asked and when) in the case you mention, the referendum question was asked third after two leading questions designed to get participants thinking in a certain way and/or the lack of preamble to the referendum question (i,e, it was just the actual question on the day): all the pollsters, to some extent or another use preambles to the question which can have an effect (given the bun fight over the actual question to be asked, you can see why YouGov's 'leaving the UK' preamble, alongside 'if the vote were held tomorrow' could shape particpants thoughts). Nevertheless, IPSOS are the outlier now, ICM, Survation and to a lesser extent YouGov, are all far closer to Panelbase than to Ipsos now so the actual No lead is probably more like the 11-12% in those polls than the 25% seen in the Ipsos poll.

Really depends which pollster you want to believe, Ipsos remain firmly wedded towards the No friendly category, with ICM and Panelbase tending towards the other, with yes at least 5 points higher and No a good 10 points lower. ICM changed their weightings over their last three polls, and you just had a YouGov one out that showed no dent in the Yes or No numbers as a result of his 'intervention'. You also have evidence from at least one poll carreid out by Scotpulse (could be cowboys) that showed very little movement from the Osborne announcement and what there was, slightly the better for yes. Ipsos show much lower levels of undecideds than the other, and it would be nice to know how that 44% who hadn't changed their position were leaning previously. Ultimately, if Ipsos are showing the actual situation on the ground then the game's a bogie (even if the undecideds all broke for Yes, on Ipsos' largely static polling numbers there'd still be a 20 point odd deficit to overcome, if it's Panelbase and ICM, then it's all to play for. methodology is everything.

I did my degree in Electronics in Heriot Watt, but I've got colleagues who went to both Glasgow and Strathclyde. Really, it's about what flavour of electrical engineering you are after. It's an incredibly wide discipline - Strathclyde are really good for the heavy engineering stuff, High Voltage stuff, power distrbution networks (they should be, they poached most of their high voltage staff from HW) Glasgow Uni is better at the light current, microchip stuff.

Actually, it's not just the one guy being out, it's the fact that we haven't managed to put a settled back four on the pitch at any time, due both to injuries and suspensions. Laidlaw's form coming in for McGurn was patchy and he did fluff a few goals for us, but it certainly didn't help going into three games with three different central defensive partnerships in front of him we had both cardle and Anderson out for two games which stifled our play as well, you also had Fox playing through an injury for several weeks due to simply not having any other options but for him to play.

If you take the Scot goes Pop poll of polls, you can see a pretty clear downward trend in No, with yes chipping away at that. Assuming a completely linear continuing trend then yes would have reduced the No lead to 5-7 points across all polls come polling day: Including Don't knows, given the healthy number of DKs seen in the polls to date, this means it should be possible for yes to win it, but it would require taking a large majority of the present DKs, or increasing the rate at which voters are converted (presently you are looking at an average of 1.02% per month, basically you need double that). edited to add: Not accounting for turnout, of course. If we assume that yes voters are more likely to hit the polls, then any turnout below 100% will give Yes a marginal boost, but how low would turnout ahve to be for that boost to be significant? edited to add: said Exclude when meant include. Becuase it's early in the morning.

That's not what I took from it, his caveats regarding methodology and age group sampling applied equally to the Survation poll that showed the +6 to yes swing. If you change your sample weighting, you will get a different result, the extent to which methodology changes in both Survation and ICM are significant is up for debate, and could be masking some shift in either direction. In addition you have panelbase showing a -2 swing from No (with Yes n/c) and also the Scotpulse poll showing a dead heat between Yes/No when undecideds are excluded and within that last group, 5% more likely to vote yes after the Osborne intervention and 3% more likely to vote no (obviously, this mob might be a bunch of cowboys).Both Survation and ICM were carried out after Osborne's foray north. I think probably the fairest thing to say is that Osborne's intervention was not the game changer he thought it would be, or the knock out blow I thought it would be (although it still might, it could be a slow burner). Curtice's take on it: http://blog.whatscotlandthinks.org/2014/02/second-post-currency-row-poll-still-no-clear-impact/

http://scotgoespop.blogspot.co.uk/2014/02/new-panelbase-poll-shows-pro.html

http://www.timeshighereducation.co.uk/news/physicists-are-more-intelligent-than-social-scientists-paper-says/2011442.article Should've read:

True, so there are many permutations on what the effect of Osborne's intervention might have been in that case - the most you can probably say is that the intervention has had no appreciable negative effect within Survation's polling sample (even if the entire positive swing were due to changes in methodology)

Where did you get your numbers from, they seem a little out of date: last time I saw we were up to 38 million in Seville.

No, the Rangers support will be coming out something ridiculous like 90% in faovur of No with their reasoning being a strong emotional attachment to the Union and it's institutions, whereas the 'mainstream' is showing anything from a 15 point to 4 point spread in faovur of No, where the arguments are far more raitonal and hard headed. The Rangers support is as much an outlier as anything.

It's not a time series because it's collected as such, it's time dependent becuase voter opinion can change inside of the sampling interval he uses, you are correct of course that randomisation might by chance land him with a rock steady sample, but at the same time, in a live campaign, and with other pollsters showing variances within the sampling period he uses, he is leaving himself open to the chance of aliasing as his sample might be varying during the polling event. Were he to conduct his fieldwork over a shorter period of time, there would be less chance of that happening. As it stands, there is the opportunity for a statistically signifcant portion of his sample to change their minds after he's polled them but before his sampling period is over, and to be counted one way, when in fact, this is an innacurate representation of their views.

That's kinda the point, the sampling interval is long, and can lead to aliasing of information, therefore the position of the population over that period of time could be innacurate. Other polling shows variances within the sample period chosen by Curtice, if the interval is too long, then these variances are missed.

Eh, we aren't?

There is a dependency on time, all polling relies on a consistency of opinion from the start to finish of the sampling period, 1-2 week fieldwork is common and has a reasonably small risk, but 3 months? Huge risk of people asked in June not holding the same opinion in October, and therefore a larger risk of his data not being a true reflection of the state of play, I understand his methodology is consistent over the years and works fine with social issues, but as soon as this became a live campaign, then the amount of information being put into the public domain for me means that his sample period is too long - Ipsos Mori showed a 5% swing away from No to Undecided in a quarter, Curtice's methodology would have missed that, or at least seen an innacurate image of it.

The sample here has a clear dependency on time. In this case Curtice's sample period is so long that he can't but help end up aliasing his own data (that's an engineering term by the way).

There is definitely a lot to use and abuse for both sides of the debate there, as Wee Bully points out the much trumpeted all time low 23% figure is now at 29%, so a six point jump there. The No side will trumpet the absolute unsailability of that headline figure. It's a bit of a contradiciton, but while that 23-29% figure exists, so does one question that reckons that all decisions should be made by the Scottish parliament has dropped by 4% since last year, which doesn't seem self consistent. My two big complaints with Curtice are as always about timing. You simply can't treat a poll of 1,500 folk over six months as one sample. I'm not the expert here, Curtice is - but in my job I work with statistics a fair bit in terms of designing semiconductor devices for quality and reliability. I know that if I tried to look at lifetime testing of parts in six subsamples over a period of six months, and then presented the data as one sample with the end points normalised to the last batch, I'd be laughed out of the building. Curtice's methodology does not allow for folk changing their minds during his sample period, he has no way of knowing that the folk he asked in June, still hold the same opinion in October. It also means that each sub-sample taken on it's own, if it can be broken down that way, would be very small and prone to large errors. Secondly, the press are reporting this as being 'new' i.e. current - when in fact the last samples of data were picked up in Ocotber 2013: Three months ago. Coincidentally, one month before the publishing of the white paper. Whatever it's actual impact, there is no doubt that there is now a lot more information available from both sides, and a lot of ink spilled since Curtice finished his data collection, which surely pours some doubt on the questions he asked regading how informed voters felt.

Why is there no reason to suspect that DKs will not split differently from the current Yes/No? Why is your assumption that this will be so of more relative worth than a private poll (the ambiguity of which I made clear when I posted the link to it, I am not so precious that I cannot take a detatched view of these things). If you look at the DKs from the WoS/Panelbase poll it showed that the majority of DKs come from Labour/SNP - the LDs and Tories are pretty much dead set against. The STV/Ipsos-Mori poll in December showed the No vote lead decreasing from 19% to14%, with the Yes vote steady -i.e. the DKs were swelled by a group of people who were travelling from No to DK, if not convinced by independence it at leaset shows a measure of traffic in that direction. I would like to know a lot more about where the DKs ar ein relation to the debate right now, The limited data suggests some positive momentum for yes in there, even if you don't take jenkin's assertion from his internal polling at face value - and I don't, I'd want to see the methodology, and am happy to examine each poll on it's own merits. You have provided no data for your assumption that DKs will break the same as current polling suggests those who have already decided will (and to look at Ipsos-Mori, even that ratio is changeable), and for all your claims that Jenkin's assertions are bullshit, it's not unreasonable to suggest that your own 3:2 against in the DK bracket should be filed under the same heading.

The point is that you have no data on the undecideds, and for all the internal polling used by the SNP may be of dubious methodology, it's still probably better than you airily pulling a ratio out of your arse, isn't it?

So not going on "actual polling" then. Fair enough.

So which poll showed the undecided's breaking 3:2 against?