Tags
(not satire – it’s the Lib Dems!)
This is extraordinary.
According to Lord Ashcroft‘s latest polling, more Liberal Democrat voters would prefer Tory leader David Cameron as prime minister than their own leader:
.
Proof – if any more were needed – that the Liberal Democrats are now much more Tory-lite than Liberal.
.
Please feel free to comment.
penniewoodfall said:
In hopeless naivety…I am voting for ‘none of the above!’
LikeLike
FinkFurst said:
Tom – FFS look at who conducted the poll.
“Proof – if any more were needed” – You have no idea what proof is……or you’re not interested.
LikeLike
GGD said:
What is he willing to do to hang onto his ministerial car?
LikeLike
Tom Pride said:
You – like a lot of people – have no idea about polling.
Who commissioned the polling has no effect on the results.
LikeLike
FinkFurst said:
“Who commissioned the polling has no effect on the results”
HA HA HA HA HA!
LikeLiked by 1 person
beastrabban said:
Reblogged this on Beastrabban’s Weblog and commented:
The transformation from Lib Dem through Tory Lite to true Tory is almost complete, just like Napoleon and the pigs at the end of Animal Farm.
LikeLike
Pete B said:
The LibDems will be finished has a political force in the UK after 2015,for the foreseeable future.
They campaigned on a Left Winged platform,but chose to be right winged.Not so much,a few white lies.More a Con Trick.
LikeLike
Tom Pride said:
Polling companies are businesses – who make their money from corporate surveys – not politics.
If you were a large corporation which wanted a polling company to find out for you what prospective consumers want – are you going to use one which tells you the truth – or one which tells you what you want to hear?
Not only do you know nothing about satire – you know nothing about PR and polling either.
LikeLike
Terry Kinnard said:
Tom – If a client is trying to find oout what people think your answer MAY be correct but a statistic is only of any use if you completely trust the methods used and the raw data that stats are based on. In either case it is easy to manipulate or specify tha answer you want by skewing the wording of the question or the group of people used to answer the question I think you are being way too trusting to believe that the group of respondants used is guaranteed to be a representative cross section of the general voteing public, however much Lord Ashcroft or the pollsters profess that they are.
[Polling companies – including those used by Lord Ashcroft – have to adhere to the extremely strict guidelines laid down by the British Polling Council] – TOM
LikeLike
FinkFurst said:
Tom – Maybe you need to give your ego a rest, ask a question occasionally, and maybe you’ll learn something…
In fact I’m a pretty experienced statistician, from my job as a biology/molecular biology research scientist. Let’s take the very first requirement for an objective poll… Were the interviewees selected randomly? As far as I can see Ashcroft makes NO such claim (in this poll or any others) so the chances that they WERE selected randomly is pretty much nil.
[Like I say – it’s clear you know nothing about polling. Interviewees in polls are not supposed to be random – they are supposed to be representative.
Read about weighting in polling – a very complex subject – and when you’ve learnt something about it come back to me.
I suggest you start your research here; http://www.britishpollingcouncil.org/pdf/andrew-cooper.pdf and here: http://race42016.com/2012/04/26/polling-101-weighting-the-sample/] – TOM
LikeLike
FinkFurst said:
“If you were a large corporation which wanted a polling company to find out for you what prospective consumers want – are you going to use one which tells you the truth – or one which tells you what you want to hear?”
Tom – I find it strange that in some ways you seem to be very experienced and cynical, and in some ways a bit naive. In my career in academia and industry, when I have come across that, it usually means there is a hidden agenda behind the apparently naive comments.
It would help if you told everybody plainly if you are a committed Labour Party activist or not……
LikeLike
NoOppression said:
Shame on the Liberal Democrats For having been Lobby Fodder For the
Tory Millionaire Supremacists
Shame on Them
LikeLike
david mellor said:
i want to toss you off.
LikeLike
FinkFurst said:
SurveyMonkey???? PLEASE tell me that you’re being sarcastic! I think I might possibly have heard about weighting maybe once or twice before. Do you know how to weight for a factor which is unknown? Of course you don’t, and nor does anybody else!
Even if you weight for the very simplest ~50/50 sex factor (as in your link) election turnout by sex is NOT 50/50, so it doesn’t work. You can choose when and how to weight according to your motivation, and most people would call it fiddling the results. GROW UP AND LEARN!
[Your comments here show you are about as ignorant as it is possible to be about polling. I feel a bit like an economist must feel trying to explain to someone really thick that prices can rise even while inflation is going down] – TOM
LikeLike
FinkFurst said:
“Interviewees in polls are not supposed to be random – they are supposed to be representative.”
…and the only way to truly represent a population from a sample is randomly. It’s not easy in some situations, but if a study doesn’t even say how it took the sample, or weighted findings thereafter, then it’s not worth reading.
LikeLiked by 1 person
FinkFurst said:
…and now you’ve changed the links you provided. Pathetic!
P.S. Read your new links yourself! http://race42016.com/2012/04/26/polling-101-weighting-the-sample/ is actually explaining about how the polling company can use ‘weighting’ to pretend that the sample is representative when it’s actually far too small to be representative, and what’s really needed is a larger random sample.
[FinkFurst, your ignorance is mindblowing. You are behind the times when it comes to psephology by around 80 years (1936 to be exact) Read this to show how ignorant your comment really is; http://historymatters.gmu.edu/d/5168/%5D – TOM
LikeLike
Strangely said:
Conspirational NHS splitter from the 80s and profiteer from the cleaning services his “polls” at the time demonised,,,,, Ashcroft, him of uncertain residence for tax purposes etc etc etc – well, his poll data is here:
Click to access ANP-140721-Full-data-tables.pdf
On some questions, only half or 3/4 of the 1007 even answered.
It “says”:
1007 sample size
It “appears” to be weighted by age and sex.
The questions were asked to phone owners, half being mobile – So no-one left the office and folks without phones don’t count.
Not very representational start is it?
Do I trust the survey?
Not a jot.
Ashcrofts history (along with Maud, Chope and Pickles) during Thatcher’s time show fixed figures proving that firefighters were over-paid etc leading to the current state in London of firemen being forced to work into old age else losing their rightful pensions, plus Boris has bought tenders to hose anyone who protests.
Don’t forget the fake reports which ending up privatising cleaning (for Ashcroft, naturally, to benefit) and the later needless deaths from MRSA and the like due to poor cleaning….
It’s highly likely that the figures are fixed IMHO. I don’t believe a single word from the creep and his cronying party.
Just because the figures are there in print does not mean that they’re genuine – who monitors it all?
A. Probably no-one, despite Tom’s protestations about surveys working for clients – the only client Ashcroft has is the tory party – just look at the categories in his “poll” website…. all politically related – nothing commercial in it at all.
LikeLike
Thomas M said:
The Lib Dems seem to be losing their own political identity.
LikeLike
Tom Pride said:
@Strangely – there are a lot of criticisms to be made about Ashcroft – as you’ve mentioned in your comments above – but his polling isn’t one of them.
ALL polling companies do commercial surveys. It’s how they make their money. Political polls are a kind of PR exercise for them. If a polling company gets close to the actual election outcomes – they get more commercial clients. That’s why polling companies ALL try very very hard to get the right results.
Your criticisms of the actual polling methodology shows the same widespread ignorance of polling that a lot of people have.
For a really good explanation of what people get wrong about polling – exactly as you’ve demonstrated in your comments above – have a look at this:
http://ukpollingreport.co.uk/blog/archives/4874
Ashcroft does not poll for the Tory party. He did – before the last election. But he was sidelined by Cameron as the main Tory election strategist.
As someone who really really doesn’t want to see the Tories win the next election, this is something I’m extremely happy about. IMO the Tories would be doing much better if they were still using him as a strategist.
I’m a bit of a polling geek. And I can assure you Ashcroft’s polling is above board. It’s also interesting to note that he releases the data for all to see. This is not something that would happen if it was for the benefit of his own party.
In fact, I expect there are a lot of Tory party strategists at Conservative Party HQ cursing him for that. Perhaps it’s his revenge for being sidelined.
Those who would like to see the Tories lose the next election would do well to ignore the fact he’s a Tory lord and pay attention to the extremely useful data he is producing. Especially as regards his polling of marginals.
Coincidentally, Ashcroft has produced a very interesting and extremely detailed poll of Tory/Labour marginals today. It shows a 5.5% swing from Con to Lab in the most marginal seats. That is very bad news for the Tory Party:
http://ukpollingreport.co.uk/blog/archives/8904
LikeLike
FinkFurst said:
“I’m a bit of a polling geek. And I can assure you Ashcroft’s polling is above board. It’s also interesting to note that he releases the data for all to see.”
He releases SOME data for all to see. Are you really that gullible?
[No. Just intelligent and informed. Under British Polling Council rules all data from a poll must be released. Besides – as illustrated by today’s polling of marginals – most of the data he has released is damaging to the Tory Party.
With every comment you make FinkFurst, my estimation of your intelligence takes a nosedive.] – TOM
LikeLike
Tom Pride said:
It amuses me immensely that of all the criticisms of this blog post nobody was clever enough to pull me up on an actual polling ‘mistake’ I did make.
‘Mistake’ is in inverted commas because I must admit I was aware of it when I wrote the post.
I’ll give a clue to anyone clever enough to genuinely debate polling with me – the poll I’ve quoted has a Margin of Error of around 3%
LikeLike
ackie00 said:
That means the Lib Dem voter’s preference for Cameron over Clegg is not statistically significant – one could argue they’d (on average) be just as happy with either.
LikeLike
Tom Pride said:
Exactly. However – while not being so headline making – the fact that Lib Dem voters are evenly divided between Cameron and Clegg is also significant. Most party voters would prefer their own leader to any other party leader. So why don’t those Lib Dem voters who prefer Cameron just vote Tory?
Lib Dem voters really are a strange lot.
LikeLiked by 1 person
FinkFurst said:
The USA Literary Digest in 1936 obviously didn’t randomly select their sample either!!!!!! They don’t even claim to have done so. Have you got any straws left to clutch?
You haven’t got a f*cking clue about statistics…… admit it! You don’t even know what a confidence interval is……
(or at least you didn’t until you Googled it just now!)
[Yet more ignorance. The point is that the Literary Digest thought – like you – that larger numbers of respondents equals more accuracy (“what’s really needed is a larger random sample..”)
A larger sample does not necessarily produce more accuracy – especially if it is not weighted properly. Voters are not a random cross-section of society. Correct weighting produces more accuracy. With correct weighting a small sample is enough. This was realised way back in the 1930s. The science is really extremely sophisticated these days – and surprisingly accurate. You need to catch up.] – TOM
LikeLike
FinkFurst said:
“Voters are not a random cross-section of society”. You really don’t get it! Voters ARE the population from which the representative sample needs to be taken in an election poll. Society is completely irrelevant!
[Exactly – now you’re starting to think. Society is irrelevant. So random samples taken from society are not going to be representative of voters. Now the question is – how to find those actual voters and make sure they are representative of the people who are actually going to vote. Random sampling will not achieve that. Only sophisticated weighting will get close to it. Of course – there is always a margin of error. These days it’s typically around 3%.] – TOM
“With correct weighting a small sample is enough.” How small? Three? You’re a f*cking idiot!
[Not three no. What childish arguments you’re making.
Read and learn (from the excellent UK Polling report archives):
“In the 1930s in the USA the Literary Digest used to do mail-in polls that really did survey millions of people, literally millions. In 1936 they sent surveys to a quarter of the entire electorate and received 2 million replies. They confidently predicted that Alf Landon would win the imminent US Presidential election with 57% of the popular vote and 370 electoral votes. George Gallup meanwhile used quota sampling to interview just a few thousand people and predicted that Landon would lose miserably to Roosevelt. In reality, Roosevelt beat Landon in a landslide, winning 61% of the vote and 523 electoral votes. Gallup was right, the Digest was wrong.
As long as it is sufficent to dampen down sample error, it isn’t the number of people that were interviewed that matters, it is how representative of the population they are. The Literary Digest interviewed millions, but they were mainly affluent people so their poll wasn’t representative. Gallup interviewed only a few thousand, but his small poll was representative, so he got it right.”] -TOM
LikeLike
FinkFurst said:
“random samples taken from society are not going to be representative of voters”
Of course they aren’t! I said absolutely nothing about society YOU DID!!! A random sample of sufficient size taken from voters will be representative of voters!
[Nope. A random sample of sufficient size will NOT be representative of those who are likely to vote. A random sample will ALWAYS be unrepresentative – unless it’s big enough to include all voters. And a survey that includes all voters is called an election! You really are not intelligent enough to understand the basics of this, are you?] – TOM
“The Literary Digest interviewed millions, but they were mainly affluent people so their poll wasn’t representative.”
PRECISELY. IT WASN’T F*CKING RANDOM, THAT’S THE WHOLE F*CKING POINT I’M MAKING!!!!!!!!!!!!!!!!!!!!!!! DO YOU EVEN KNOW WHAT RANDOM MEANS?
You’re trying to say that you CAN get representative results by ‘intelligent’ weighting even if your sample is small or NOT random! I’m saying it has to be random AND large enough. Are you really too thick to understand that?
[I’m not trying to say anything. I’m pointing out that relatively accurate polling (MoE of 3%) has been done using small, weighted groups for the best part of 80 years.
This is like arguing with someone who still doesn’t believe TV is possible.
Strangely, for someone who claims to be a statistician, you seem to be unable to understand the meaning of ‘random’. Random sampling doesn’t guarantee anything useful in polling of voting intention. I could take a random sample of red and blue socks and be just as likely to get nearly all red, nearly all blue or any mix of the two. A larger sample will just guarantee the sample to be more likely to be an even mix. But in political opinion polling, actual voters are not equally distributed like socks. In fact, they are extremely unequally distributed amongst society (or eligible voters if you prefer). You see – an opinion poll will survey anybody who is eligible to vote. But not everybody who is eligible to vote will actually vote. And interestingly, asking people how likely they are to vote is an infamously very unreliable way of finding out whether they will actually vote or not. So random sampling will ALWAYS be inaccurate when surveying voting intention.
Are you sure you’re a statistician?] – TOM
LikeLike
FinkFurst said:
If you have ANY idea whatsoever, then you should be able to explain the MOST basic statistical concept of a normal distribution. Can you?
LikeLike
FinkFurst said:
There’s little point in keeping on saying that you’re right and I’m too stupid to understand. So let’s put you to the test Mr Polling Expert – Please explain how Ashcroft selected the 1,007 telephone interviewees in order to make them representative of all adults in Great Britain…
LikeLike
FinkFurst said:
Also lets try another question to see if you really understand – Assuming that all other aspects of a poll are the same, would you agree that the larger the sample the more likely it is to be representative of a population?
LikeLike
FinkFurst said:
P.S.
“Polling companies – including those used by Lord Ashcroft – have to adhere to the extremely strict guidelines laid down by the British Polling Council – TOM”
As far as I can see Lord Ashcroft Polls is NOT a member of the BPC, so no it doesn’t.
LikeLike
overburdenddonkey said:
well it’s good to see global ills being righted….
LikeLike
Strangely said:
Ashcroft does not need to make money from polling…. he’s well rich enough and continues along his own tory agenda whether he’s officially part of it all, or not.
Therefore, huge doubts can be cast over his “polling” – there are no other polls on his polling website except politically motivated ones. Only summary, not detailed data is shown. Despite what you’ve said Tom, I can’t find Ashcroft as being part of the BPC – http://www.britishpollingcouncil.org/officers-members/
To say his polls show the tories in a bad light is just wrong. His slant and commentary make this plain to see.
As for the accuracy of the published summary results, I note in some of the samples a standard error of 0.2, i.e. 20% and more. This relates to the error in the population, not the sample, thus the figures could be 20% out…. In other words, you could make almost anything out to be true, probability-wise.
In short, the figures could all be made up, they could be adjusted to look good-ish or be completely accurate. I can’t tell.
All I can tell is that Ashcroft has form which goes back a long way and continues on his agenda whether he’s officially in or out of tory governance and/or finance. Behind closed doors is where it happens.
(Don’t know how you get 20%. Margin of error in the polls is 3% -standard for the industry. He released the full tables for the polling – not just the summary. See them here: http://lordashcroftpolls.com/wp-content/uploads/2014/07/LORD-ASHCROFT-POLLS-Con-Lab-marginals-report-July-2014.pdf) -TOM
LikeLike
guy fawkes said:
All poll outcomes are fiddled!
LikeLike
FinkFurst said:
Mr Strangely – I agree with everything you said…… and that’s a bit unusual for me!
LikeLike
FinkFurst said:
No they aren’t. More ignorant drivel.
LikeLike
guy fawkes said:
In my opinion they are – just like statistics they are manipulated.
Who takes any notice of them anyway? your job is redundant Fink!
LikeLike
FinkFurst said:
“Who takes any notice of them anyway?”
Errrrmmm … in case you hadn’t noticed, quite a few people do.
LikeLike
Tom Pride said:
@Strangely – What an ignorant lot you all are. Ashcroft isn’t polling to make money! Quite the opposite. Ashcroft is the client. It costs him millions to have this polling done for him.
@FinkFurst – Yet more ignorance. Ashcroft Polling isn’t the company that does the polling – that’s why it’s not a member of the British Polling Council. It’s like expecting a doctor’s patient to be a member of the BMA!
Ignorance ignorance ignorance.
I’ll leave you all to wallow in it. I’m out.
LikeLike
Tom Pride said:
IGNORance. The interviewees were not selected to be representative. They were WEIGHTED to be representative.
Please learn the meaning of weighting in psephology before we go on with this conversation.
LikeLike
Tom Pride said:
(Facepalm and a big sigh!) It absolutely shouldn’t be representative of the population! It should be representative of the actual VOTERS who will actually vote in an election – who are NOT representative of the population (as I’ve explained a hundred times to you before). Are you thick? The people who actually vote in an election are NOT representative of the population. Isn’t that obvious?
LikeLike
guy fawkes said:
Ashcroft didn’t make his money through relying on Polls.
If he is paying to have this polling done for him then he has more money than sense.
LikeLike
FinkFurst said:
Ashcroft Polling publishes the findings, so the rules would apply to him as well as the company which conducted the poll.
Did you know that Ashcroft was invited to join the BPC and has not (so far) agreed to do so? Apparently you didn’t!
LikeLike
FinkFurst said:
(Facepalm and a big sigh back to you!!!)
I didn’t say THE population, I said A population. It’s the one most basic statistical concept there is!!!! Just put your ego away and read this – http://www.statistics.com/glossary&term_id=812
LikeLike
FinkFurst said:
Wrong, but please explain how were they selected…
LikeLike
FinkFurst said:
Tom – When you’ve finally come up to speed with Day 1 of GCSE Statistics, perhaps you could answer the question.
This might also help – http://www.edexcel.com/migrationdocuments/GCSE%20New%20GCSE/GCSE_Statistics_Spec_2012.pdf
LikeLike
FinkFurst said:
Tom – I just realised… have you been under the same misapprehension about the meaning of the word “population” in every statistical study you’ve ever seen? No wonder you were confused and kept talking about society!
LikeLike
FinkFurst said:
Tom – By the way, if you read the BPC rules you will find that they apply only to whoever puts the findings into the public domain (see rule 2.1). So that IS Ashcroft, which is not a member. I hope that fact puts this point to bed now…
LikeLike
Strangely said:
@Tom – you said, “Ashcroft isn’t polling to make money! Quite the opposite. Ashcroft is the client. It costs him millions to have this polling done for him.”
Quote from Ashcroft Polls dot Com on the about page:
“Welcome to LordAshcroftPolls.com. The site is a comprehensive guide to my polling work, and my commentary on politics and public opinion. It is intended to give full details of my research while being as accessible as possible.”
So:
“my polling work”
“my commentary on politics”
“intended to give full details of my research”
My, My, My.
It IS his site and his work. He does it to sway public opinion. He says so.
@Tom – I said standard error which is what’s on (say) page 1 of the so-called detailed data:
Click to access ANP-140721-Full-data-tables.pdf
I’m not a statistician so I had to look it up as well as the proper definition for standard deviation.
The standard errors where shown hang around 0.1 with some over 0.2. 0.2 equates to 20% since 1 is top. Wikipedia says,
” standard error of the sample is an estimate of how far the sample mean is likely to be from the population mean, whereas the standard deviation of the sample is the degree to which individuals within the sample differ from the sample mean”.
To me, this sounds like the figures can mean anything to anyone.
It also means that since Ashcroft’s objective appears to get people noticing his figures bollocks, then he’s won.
That’s all it was ever meant to do. That is what he does.
These links below, relating to PULSE and stuff around it, show that Ashcroft’s game plan and method of playing it have not changed in 30 years… It’s all to do with grabbing state stuff for the benefit of him and his ilk.
Quote:
“A confidential memo on PULSE’s campaign launch says its main objective was to “increase public awareness of the advantages of using private firms to provide local and health authority services””
Since that time this has expanded to the point where unqualified teachers are employed in so-called “academies” and privatised firms have been bailed out, compensated for failing and given new contracts AFTER committing fraud against the state and taxpayer.
Ashcroft’s game plan, (paraphrasing another tory), thrives on the oxygen of publicity.
I hope you don’t mistakenly give him more free screen time.
These are 3 pieces on PULSE and the connections. Many names are familiar and now doing exactly the stuff they planned over 30 years ago.
http://www.powerbase.info/index.php/Michael_Forsyth_Associates
http://www.1in12.com/publications/library/pickles/chap6.htm
http://www.highbeam.com/doc/1P2-18710009.html
[Of course Ashcroft calls it MY research. He pays a LOT of money for it.] -TOM
LikeLike
FinkFurst said:
Tom – [Of course Ashcroft calls it MY research. He pays a LOT of money for it.]
…and you don’t seem to be even vaguely curious about why! Why do you sometimes seem to be so naive?
Strangely – The data as presented by Ashcroft are very unusual and the analysis details are so limited as to be incomprehensible. High standard deviation/standard error essentially means that the data as collected are too variable to draw confident (or any) conclusions. The usual answer is to take a larger sample from the population (no Tom, that doesn’t mean the population of the country!)
You mention the details on page 1 (Q1), but if you look further down you will find that NONE of the subsequent questions show any error analysis whatsoever!
LikeLike
FinkFurst said:
Tom – Do you understand what “population” means now?
LikeLike
Tom Pride said:
FinkFurst – this is my last comment to you. As I have explained to you countless times – population in this case is the people who ARE GOING TO VOTE. And as I’ve also explained to you countless times – that is not the same thing as the population who are eligible to vote. So – as I’ve explained countless to you countless times before – you can’t find those people and balance the sample in terms of party voting intention by random sampling. Random sampling CAN’T do that. Weighting not only can do that successfully but it has been doing that for over 80 years.
You are trying to disprove science which was proven decades ago.
Every time I try to engage you in intelligent debate you resort to insults. In this one thread you’ve called me a liar, dishonest, a wimp and countless other childish insults. Now why don’t you do us all a favour and STFU!
LikeLike
FinkFurst said:
Now you understand what “population” means in statistics, I’ll try asking my questions again:
1) Assuming that all other aspects of a poll are the same, would you agree that the larger the sample the more likely it is to be representative of a population?
2) Please explain how Ashcroft selected the 1,007 telephone interviewees in order to make them representative of all adults in Great Britain…
P.S. We still haven’t discussed “Satire is ALWAYS partisan”. Are you up for that?
LikeLike
Tom Pride said:
1) Only if it’s 100% of the population. The population in this case is hidden ie – the people who are going to vote at an election. You can’t find that population unless you weight the sample in a very sophisticated way.
2) He didn’t select them to represent all adults in GB. That’s pointless. He needed the sample to be representative of all people WHO ARE GOING TO VOTE in GB. That’s impossible to select. So – as with all market research companies – they were WEIGHTED to make them representative of all adults WHO ARE LIKELY TO VOTE in GB.
3) Ashcroft didn’t select anyone. A polling company did. He’s just the client.
4) Any more childish insults and I’m out. Keep the discussion civil and intelligent and we can continue.
By the way – market research – and particularly political polling – is different from most statistical sampling. I’ll give you an example. When pollsters ask for example – who are you going to vote for – a certain % of people say “I don’t know”.
What do you do with those people? Exclude them from the sample? What if – as it does in fact turn out to be the case – one particular party has more voters who like to say “I don’t know” when asked voting intention but in fact are still likely to vote for that party?
If you exclude them – one party will register lower voting intention than it should and your sample will not be representative.
Pollsters – which are actually just market research companies – are very good at dealing with those kinds of problems. They’re good at it because commercial companies will pay very well for MR companies to conduct their market research – if they can prove they can deal with those kinds of problems.
LikeLike
FinkFurst said:
1) So if the population is ~1 million, a sample of 2 is just as likely to be representative as a sample to 20,000? Is that seriously what you think?
2) So how WERE the interviewees selected?
3) True, but obviously I wasn’t implying that Lord Ashcroft selected and ‘phoned them himself!
4) I only ever use an insult when I can justify it. Challenge me any time.
“political polling – is different from most statistical sampling” – No, it’s NOT different. To use your own example, if you ask surgery patients what their level of pain is, or supermarket customers if they like the new trolleys, some will say they don’t know.
[Yes – I think 2 is a good enough sample (that was written in sarcasm font.)
I warned you to debate intelligently. That’s childish. Any more of that and I’m out.
Commercial companies really need to know if it really hurts or not. For example, if coca cola wants to know what people really think of a new drink they need to do market research. And if a % of people who say “I don’t know” really hate it – coca cola want to know that, not just ignore it. In political polling we get to find out for real what people will vote for so it’s a good PR exercise for an MR company to show they can get to the real truth of what people think] -TOM
LikeLike
FinkFurst said:
Oh come on!!!! You just clearly said that a larger sample size is only more representative if it’s 100% of the population! So which is better – 2, 20, 200, 2000 or 20,000? Or are you saying that they’re all equally representative?
How your answer to question 2)?
LikeLike
Tom Pride said:
If you are really interested in the technicalities of polling and Ashcroft’s in particular – and not just trying to have a go at me – can I suggest you direct your questions to the people on the UKPollingReport forum. They are from all political persuasions but are strictly non-partisan in their discussions and a finer set of polling geeks won’t be found anywhere else.
They will tell you in great detail how interviewees were selected, exactly how don’t knows were realllocated, what was weighted by how much etc etc.
Mind you, they’ll give you short shrift if you ask them silly questions like you’re doing with me.
Ask your questions here:
http://ukpollingreport.co.uk/blog/archives/8904
LikeLike
FinkFurst said:
I’ll try the questions in a different way and maybe I’ll have better luck!:
1) In a statistical study of a population of ~1 million, is a sample size of 20,000 more representative than a sample size of 20? – Yes, No, Don’t Know.
[As I keep explaining to you, NO. Because if it’s 20,000 people who aren’t going to vote it’s as useless as a sample as 20 or 2. You can keep asking me the same question and I’ll keep giving you the same answer. So stop – because it’s boring!] – TOM
2) Can you find any reference to how the company employed by Ashcroft Polling in the above noted study selected the 1,007 telephone interviewees – Yes, No, Don’t Know
[Polling companies select their interviewees in a few different ways. As I’ve already mentioned – if you’re really interested, ask the experts at UKPollingReport. They’ll tell you everything about it.] – TOM
LikeLike
FinkFurst said:
1) You ACTUALLY think asking 20 people will give you the same result confidence as asking 20,000 people??? Of course it’s no problem if that’s what you think, but you will be judged accordingly.
2) Presumably you read the report before you wrote this blogpost, and the question is – Can YOU find it in the report? After all, YOU said that it contains proof. Go on… try again – Yes, No, Don’t Know.
LikeLike
FinkFurst said:
1) I think you’ve proved that you don’t understand statistical principles, so there’s no point flogging that horse any more.
2) Let’s look at the politics of this question – We’ve established that Ashcroft hasn’t published any methodology of interviewee selection or weighting, nor is he required to do so by the BPC because he has declined to join it. Would you therefore entertain the possibility that he COULD have used methods which would engineer a politically biased result?
LikeLike
FinkFurst said:
[As I keep explaining to you, NO. Because if it’s 20,000 people who aren’t going to vote it’s as useless as a sample as 20 or 2. You can keep asking me the same question and I’ll keep giving you the same answer. So stop – because it’s boring!]
I thought I would also answer this, because it actually highlights the issue of sample size, and it might help you to understand a little about statistics. So I’ll flog this horse once more!
If you’re asking people about voting intention and your sample size is only 20, then it’s entirely possible that all 20 never vote, so your poll told you nothing about the voting intention of that population. However, you would have to be pretty f*cking unlucky to get 20,000 who never vote! Do you understand that? It’s really the whole basis of statistics…
LikeLike
Tom Pride said:
Yes but – as I’ve explained countless times – the amount of people in that 20,000 who are able to vote will not be representative of the people who DO vote. Not even close to it. Increasing the size of the sample will not solve that problem. What if there are several thousand UKIP voters who are likely to vote in the sample and just a few hundred Labour voters? Or in a larger sample a few hundred thousand UKIP voters and a few thousand Labour voters? The error is exactly the same.
Those are all useless samples – in fact ANY size sample is just as useless (except 100% of course). Because we don;t know who is actually going to vote – and how they’ll vote – on the day.
That’s why the samples ALWAYS need to be weighted to likelihood to vote, past voting intention, even what newspaper they read sometimes. If you want to know more about weighting, ask the experts on UKPolling Report. They will describe it to you in greater detail (and more expertly than I can) if you’re interested.
All a larger random sample would do is reduce the margin of error by a tiny amount – half a % point or so.
In fact, a sample of a couple of thousand is all that’s needed with good weighting. Surveying tens of thousands is time consuming, expensive and doesn’t necessarily make the survey more accurate.
In market research it’s all about weighting weighting weighting..
Most people who are ignorant of polling – including you it seems – seem to regard weighting raw data as ‘fiddling’ or ‘pretending’ as you’ve already described it. It’s not. It’s science (and a bit of an art too).
LikeLike
FinkFurst said:
You’ve now admitted that sample size IS important, in a rather strange, oblique way… but I guess that’s your character! Sometimes 1000 is enough, sometimes not, it depends on the anticipated incidence and variability in the population in question. If you’re interested in statistical significance in a population where the incidence of the point of interest is perhaps 0.1% (such as Respect voters!) then a sample size of 1000 is NOT sufficient. That was why I asked the question several times. I suspect you still don’t really understand it, but never mind.
You’ve got to the point which I was alluding to. The issue isn’t that a smaller sample size is better (or even as good), it’s because it’s CHEAPER. The larger the sample size, the less need there is for subjective weighting, but weighting costs far less than taking an adequate objective sample in the first place, so that’s why it’s done. I have written statistical software which uses weighting, so I’m not unaware of the concept. However, I ALWAYS published exactly how and why the weighting was implemented. If you don’t publish your methods (as Ashcroft) then it IS ‘fiddling’ or ‘pretending’. Perhaps it’s best if we leave statistics there for now.
How about answering my question about the politics from above – “…Would you therefore entertain the possibility that he COULD have used methods which would engineer a politically biased result?”
LikeLike
Tom Pride said:
No – as I’ve explained countless times – size is not important. WEIGHTING is much more important.
It is impossible to take random samples which will reflect real voting intention. Larger random samples will give you a better balance of the electorate – NOT THE PEOPLE WHO WILL ACTUALLY VOTE in elections.
However, I see you’ve dropped the word random in your arguments. You’ve changed your argument because you lost it.
You originally argued that larger RANDOM samples produce more accuracy in pollling. You seem to now accept that random sampling does not produce more accurate results.
Just a simple “thank you for teaching me something I didn’t know Tom” will suffice.
However, this is an example why i won’t debate with you any more. You are not interested in real debate. You are only interested in proving yourself right and others wrong. That’s why you resort to childish insults at the drop of a hat and ignore intelligent arguments. When you asked me a question whether I was a Labour party activist – you already thought you knew the answer. So when i (truthfully) told you I wasn’t – you simply called me a liar. That means your question wasn’t genuine. You didn’t really want to know the answer. It was simply a device to attack me – if I had ignored your answer – you would have attacked me for ignoring it. If I had said yes, you would have attacked me for that. And when I gave you the truthful answer – you attacked me for lying.
You are – in fact – a troll and as such – a waste of my time. I am only interested in intelligent debate here.
Anyone else want to intelligently discuss polling issues with me – I like discussing the subject so please go ahead.
LikeLike
FinkFurst said:
Oh dear! That wasn’t very edifying!
As you’ve specifically asked, I’ll try again with the same relevant question about polling issues… “Would you therefore entertain the possibility that he [Ashcroft] COULD have used methods which would engineer a politically biased result?”
LikeLike
FinkFurst said:
P.S. I haven’t dropped the concept of random sampling, and nor has the entire science of statistics!
LikeLike
FinkFurst said:
P.P.S. Actually, if you had said that you’re a Labour activist, and proud of it, then I would have said something like “Well done and good luck”. I might have challenged you about some Labour Party policies, but that’s another matter!
LikeLike
FinkFurst said:
Tom – Maybe you missed my earlier posts, so I’ll try again… “Would you therefore entertain the possibility that he [Ashcroft] could have used methods which would engineer a politically biased result?”
To put it in simple terms – Could Ashcroft be fiddling the results, and we can’t tell because he’s hiding the details?
I think that’s an intelligent question about polling issues, don’t you?
LikeLike
Tom Pride said:
No – it’s a stupid question.
What bias? Ashcroft’s polls are in line with results from all other pollsters. Which all show small Labour leads within a Margin of Error of 3%. That Labour lead has been narrowing of late. His polls of Tory/Labour marginals have also shown slightly larger Labour leads which have also been narrowing lately. Nothing unusual or pro-Tory there. In fact – not even particularly news worthy. Except for Ashcroft’s recent poll which had an interesting detail which showed Lib Dems prefer Cameron as PM to Clegg and which I duly wrote about. Nothing earth-shattering but a funny little detail for political geeks.
But you seem to prefer more interesting conspiracy theories. You seem to believe Ashcroft is in a conspiracy – which would have to include all other UK pollsters (oh and Tom Pride too) – to show a small Labour lead in order to ….. erm do what exactly? How the f*ck does consistently showing a small Labour lead help the Tory Party?
It’s silly stupid nonsense.
Believe me if there was a scandal to report – some kind of fiddling as you call it – I would be the first to write about it. As someone who hates the Tories, what motive would I have to protect a Tory Lord if he was fiddling results to favour his party? The fact is – he isn’t. Not one person from the polling world has questioned his results.
I also have to add that YouGov’s results for the Sun and the Times are often questioned by people on the Left – just because they are commissioned by pro-Tory publications. But I can assure you those results are also not fiddled. They are also above board. Not least because YouGov would lose a LOT of its commercial clients if it was found to be fiddling results.
Just because Murdoch’s paying for them doesn’t mean YouGov’s results favour the Tories. They don’t. They also show a small Labour lead of around 4-5% within an MoE of around 3%. And so does Opinium and Populus and ComRes and Ipsos Mori.
But perhaps you think they’re all in some weird conspiracy to desperately show Labour with a small lead (which is supposed to help the Tories in some obscure way which nobody can seem to explain to me).
So no – Ashcroft doesn’t hide or fiddle the results. He publishes the results and raw data the same as any other pollster does. The fiddling you refer to is normal weighting which all pollsters have to use to get accurate results (for reasons I’ve tried to explain to you endlessly and which hopefully you now understand).
I used to think you were intelligent with a good eye for bullshit.
I was wrong.
Now no more of this silliness.
LikeLike
FinkFurst said:
If polls are so unimportant politically, as you are apparently contending, why do ALL politicians watch them so avidly? Where there is such political importance, there is ALWAYS motivation for covert political influence.
I didn’t say that weighting per se is ‘fiddling’. As I mentioned, I have used weighting myself. But if the weighting methods are not published then it gives an opportunity for data manipulation. In fact the easier and less detectable way to manipulate data is to bias the original sample, or to selectively exclude some data (i.e. some of the interviewees in this case) during analysis. I have uncovered that type of manipulation several times, including in an FDA toxicology safety trial.
This brings me back to the question of why Ashcroft has declined to join the BPC. It is because he would then be required to publish his polling methods and details in full? Maybe he has another very good reason, I don’t know. I think he can afford the £750 fee!
I’m making no specific accusation as I don’t have specific evidence. I’m just saying there is clear motivation and clear opportunity, but that doesn’t necessarily prove a crime!
By the way, you might like to read this, especially the second sentence of the introduction. – http://www.ipsos-mori.com/Assets/Docs/Publications/Ops_RMC_Online_Access_Panels_08_03_10.pdf
LikeLike
FinkFurst said:
P.S. Something else just struck me from your last reply. You seem to assume that the only possibility is that people like Murdoch or Ashcroft would be trying to “help the Tory Party” or “favour the Tories”. Have you considered that their actual motivation is far more likely to be to try to manipulate the party for their own ends?
LikeLike
FinkFurst said:
Let’s look at only the “best Prime Minister” question in this poll that you’ve highlighted. There are 33% of responses which are conveniently left out of the summary. Look at the actual data tables and think about what’s NOT there…
Why did they not ask whether people thought Farage would make best Prime Minister? Or maybe they did, but that result was far too disturbing, so the responses were shoved into a “none of them” category so that it didn’t make a headline.
LikeLike
FinkFurst said:
Tom – I’ll tell you a little story. Many years ago my first job after leaving university was in a large multinational company. I was asked to do the data analysis on the safety test results of a new biological product. One of the data points fell outside the expected safe range and showed the product in a potentially bad light. I was told in no uncertain terms by my boss that in future such ‘anomalous’ data should be left out. The odd result might have been an error, or it might have been a safety problem, it was impossible to tell without doing the testing all over again, which of course would have been expensive. It later became clear that such practices were the normal culture in that company. You might be interested to know that my old boss is currently in a senior position in well known government department responsible for public safety.
Here’s a little something to look out for – If you see a graph which has a series of points at regular intervals, then a jump where one seems to be missing, ask yourself why…
LikeLike
neurontherapy said:
The statistical fisticuffs were very interesting, but overlooks the stratified sample. First take a randomised sample from your population, as large as possible, then ask them if they intend to vote, then you take a randomised sample of those who said they intend to vote, and so on.
However, the most interesting thing about Lord Ashcroft’s results, whether or not we can believe them, is that 10 Labour voters said they would like Cameron as Prime Minister and 10 said they would like Clegg. Are they stark raving bonkers or just unaware of what these two have been doing to the country for the last 4+ years?
LikeLike
FinkFurst said:
Tom – Did you read at least the introduction of the Ipsos-MORI paper? I don’t want to restate the point about sampling methods, that’s obviously pointless, I just wondered if you read it and so whether it’s a waste of my time to post such things in future.
LikeLike
FinkFurst said:
Neuron – No doubt Tom would disagree with your second sentence, because random sampling is just SO démodé!
I agree that it’s curious that somebody says they will vote for one party, but want a different party leader to be Prime Minister! Or is it just an inevitable consequence of a ridiculous ‘first past the post’ party electoral system? However, the fact that most voters haven’t really got a clue what they’re voting for is a problem which will forever plague all democracies, made worse by increasingly sophisticated party propaganda. I think there are other more interesting and concerning things about Ashcroft’s polls, but I’ve probably gone on about them enough already.
LikeLike
FinkFurst said:
…to put the principle in another way, I think Scotland should be an independent country, but not… REPEAT NOT… because I want Alex Salmond to lead it. The media and party propaganda tries to make it all about the leaders, when they are mostly irrelevant. Just look at Miliband’s acutely embarrassing ‘it’s all about me’ speech today!
LikeLike
FinkFurst said:
Tom – Did you read at least the introduction of the Ipsos-MORI paper? I don’t want to restate the point about sampling methods, that’s obviously pointless, I just wondered if you read it and so whether it’s a waste of my time to post such things in future.
LikeLike