Brexit.

[Chin Rey]

Technically offices have requirements but they are things like 'be x years old and have a law degree'

I think that U.S. only, I don't know of any other countries where lawyers domintae politics.

 

[Kara Spengler]

I think that U.S. only, I don't know of any other countries where lawyers domintae politics.

It is probably only some offices too. We have some real idiot politicians who I would not bet can tie their own shoes.

 

[Innula Zenovka]

 

[Kara Spengler]

That was just the bus too. You have to ask what will these people do when the reality of brexit hits.

 

[Luisa Land]

new opinion poll (YouGov survey) shows new figures concerning Brexit


see
The polls are clear: support for staying in the EU has rocketed | Peter Kellner


" The latest YouGov survey for the People’s Vote campaign contains three important messages.

....

1. Staying in the EU now holds a commanding lead over the government’s deal

2. Millions of 2016’s leave voters have lost faith in Brexit’s ability to make life better

3. Labour could suffer badly if it ends up facilitating Brexit"

 

[Innula Zenovka]

 

[Bartholomew Gallacher]

new opinion poll (YouGov survey) shows new figures concerning Brexit

I don't trust anything that YouGov comes up with, because their reputation in the industry is really, really bad due to sloppy methodology.

 

[Innula Zenovka]

YouGov were one of the few (the only?) polling companies to predict a hung parliament in the UK general election of 2017, it seems: How almost all the pollsters got it wrong... again

 

[Porsupah Ree]

because their reputation in the industry is really, really bad

[citation required]

 

[Luisa Land]

I don't trust anything that YouGov comes up with, because their reputation in the industry is really, really bad due to sloppy methodology.

what do you mean with "reputation in the industry"? Mostly they work for media, or is that wrong?

I am not familiar with YouGov, but a few explanations for your judgement would be helpfull
Some leading persons of YouGov are or have been members of the Conservative Party. Do you mean this causes a "bias" (Bias is disproportionate weight in favor of or against one thing, person, or group compared with another, usually in a way considered to be unfair. "
The only critique I could find were , that for Brexit surveys YouGov weigthed the ukip-voters too strong, but this they admitted themselves

"One of Britain's most reputable pollsters admitted that the methodology it used for EU referendum surveys was flawed and resulted in public support for a Brexit being overstated....
In the release, YouGov said:
We have made a slight change in our methodology since our previous poll. While our polls at the Scottish, Welsh, and London elections were generally very accurate, one consistent flaw was that we had UKIP too high.
This error appears to be related to the switch from weighting by party ID to weighting by past vote."

YouGov polls have been exaggerating the number of people who want a Brexit
(obsere that even in this critical report they write of YouGOv as a most reputable pollster)


What is the critique, what is sloppy with their Methodology? As far as I could inform me they use online-panels of people for their surveys. are the Panels ok? .Are the the interview-questions doubtfull? Are the evaluation-systems not ok?

I dont intend to defend YouGov, for that I dont know. much enough..far from knowing enough. Just interested in discussions about settings of opinion polls.

 

[Bartholomew Gallacher]

what do you mean with "reputation in the industry"? Mostly they work for media, or is that wrong?

Quite simple: their own industry, opinion research companies.

Opinion research is based on statistics; in order to get a representative view - meaning that the views from the sample you have choosen can be applied to the rest of the population - you have got to choose the sample carefully, not leaving out any kind of age/social stratum/nationality. The holy grail you want to achieve is a so called random sample. If you don't know what that such a sample is: pulling out 10 colored balls of a bucket full of 100, and predicting the colors of the rest in it with best accuracy possible.

Working correctly to get a proper sample is hard; if you fail to draw a proper sample you got what is being called a statistical bias, meaning that the statements you are meaning cannot applied to the population. A typical bias could be leaving out certain social stratums, not having a real profile in your sample of the population, and so on.

The classic example on what a bias can cause happened in the 1930s in America, the polling desaster of 1936:

the popular magazine Literary Digest asked over 10 million readers about who they are going to vote for presidency, Landon or Roosevelt, and received 2,4 million answers. Literary Digest predicted a vote of 57% for Landon and 43% for Roosevelt. So many did believe this would come true.

There was another newcomer to the industry back then, Gallup. Gallup only interviewed around 50.000 people, and predicted the real outcome of 62% for Roosevelt and 38% for Landon. What happened? Quite simple: selection and non response bias. Literary Digest sent questions to people who had phone numbers back then, and asked them to return their answers. So they asked only the middle class, but entirely left out the lower social stratums, and many people of the 10 million they wrote to just didn't answer. Gallup in return manufactured his sample far more carefully, didn't leave out the lower income classes/poor and therefore was able to get far more accurate results using much less people.

This thing is the classical example of bias, how it ruins representativity and lectured to every student of statistics around the globe in their first course.

Coming back to Yougov and their main error: they are doing their public surveys online only. So this means you have got to sign up on their web site, answer polls and earn money. They are not using any other technic which is being used by classical opionion research institutions, like phone calls, asking people directly on the street, and so on - just the internet. So they always potentially do suffer a big, selectional bias by default regardless the question they do research.

By using just the internet with this approach you do have a really heavy bias, because you are unable to reach out to all the different social stratums you need to be able to hear to get a proper sample, and therefore proper results. This is why I distrust Yougov, because this simply cannot work for many questions if not all the people you need to ask are on the internet/internet affine. Many old people for example just won't care about such stuff in the internet by my experience, but you would need to ask them directly. They are therefore weighing their samples, but this is always another big, big source for errors, because at the end of the day it is always better to have a proper random sample than to recalculate your biased one based mostly on your experience. And this possible bias is rooted in the way Yougov works, it won't go away - ever, therefore weighing their samples is for them a must. This is why I distrust them and prefer studies being done by proper, established opinion research companies over theirs any time.

Selection Bias in UK Polling (Part 2): Internet Polling
https://www.displayr.com/what-is-selection-bias/
Why opinion pollsters failed to predict overall majority for David Cameron

The only real reasons why companies like Yougov, Civey and others are having customers is because they are really cheap compared to their much older competitors; if I would ever need to do serious market research, I would never trust anything Yougov/Civey and so on might come up with.

And most of the press articles where Yougov&Co. are being praised are being rewritten press notices by Yougov, where the authors don't even have the slighest grasp about the fundamental workings of statistics/opinion research and just are taking for granted what Yougov provides for free.

 

[Chin Rey]

Opinion research is based on statistics; in order to get a representative view - meaning that the views from the sample you have choosen can be applied to the rest of the population - you have got to choose the sample carefully, not leaving out any kind of age/social stratum/nationality.

I'll probably draw some flak for mentioning this but here's a classic example how not to do a poll:
Poll shows 72% of Second Life users are more likely to promote it « Firestorm Viewer – The Phoenix Firestorm Project Inc.
It's such a shame because as far as I know, this is the only attempt ever made to do an SL users survey. It could have been so valuable but sadly the methodology had all the flaws Bartholomew mentioend in spades so the result was totally worthless.

 

[Kara Spengler]

Quite simple: their own industry, opinion research companies.

Opinion research is based on statistics; in order to get a representative view - meaning that the views from the sample you have choosen can be applied to the rest of the population - you have got to choose the sample carefully, not leaving out any kind of age/social stratum/nationality. The holy grail you want to achieve is a so called random sample. If you don't know what that such a sample is: pulling out 10 colored balls of a bucket full of 100, and predicting the colors of the rest in it with best accuracy possible.

Working correctly to get a proper sample is hard; if you fail to draw a proper sample you got what is being called a statistical bias, meaning that the statements you are meaning cannot applied to the population. A typical bias could be leaving out certain social stratums, not having a real profile in your sample of the population, and so on.

The classic example on what a bias can cause happened in the 1930s in America, the polling desaster of 1936:

the popular magazine Literary Digest asked over 10 million readers about who they are going to vote for presidency, Landon or Roosevelt, and received 2,4 million answers. Literary Digest predicted a vote of 57% for Landon and 43% for Roosevelt. So many did believe this would come true.

There was another newcomer to the industry back then, Gallup. Gallup only interviewed around 50.000 people, and predicted the real outcome of 62% for Roosevelt and 38% for Landon. What happened? Quite simple: selection and non response bias. Literary Digest sent questions to people who had phone numbers back then, and asked them to return their answers. So they asked only the middle class, but entirely left out the lower social stratums, and many people of the 10 million they wrote to just didn't answer. Gallup in return manufactured his sample far more carefully, didn't leave out the lower income classes/poor and therefore was able to get far more accurate results using much less people.

This thing is the classical example of bias, how it ruins representativity and lectured to every student of statistics around the globe in their first course.

Coming back to Yougov and their main error: they are doing their public surveys online only. So this means you have got to sign up on their web site, answer polls and earn money. They are not using any other technic which is being used by classical opionion research institutions, like phone calls, asking people directly on the street, and so on - just the internet. So they always potentially do suffer a big, selectional bias by default regardless the question they do research.

By using just the internet with this approach you do have a really heavy bias, because you are unable to reach out to all the different social stratums you need to be able to hear to get a proper sample, and therefore proper results. This is why I distrust Yougov, because this simply cannot work for many questions if not all the people you need to ask are on the internet/internet affine. Many old people for example just won't care about such stuff in the internet by my experience, but you would need to ask them directly. They are therefore weighing their samples, but this is always another big, big source for errors, because at the end of the day it is always better to have a proper random sample than to recalculate your biased one based mostly on your experience. And this possible bias is rooted in the way Yougov works, it won't go away - ever, therefore weighing their samples is for them a must. This is why I distrust them and prefer studies being done by proper, established opinion research companies over theirs any time.

Selection Bias in UK Polling (Part 2): Internet Polling
https://www.displayr.com/what-is-selection-bias/
Why opinion pollsters failed to predict overall majority for David Cameron

The only real reasons why companies like Yougov, Civey and others are having customers is because they are really cheap compared to their much older competitors; if I would ever need to do serious market research, I would never trust anything Yougov/Civey and so on might come up with.

And most of the press articles where Yougov&Co. are being praised are being rewritten press notices by Yougov, where the authors don't even have the slighest grasp about the fundamental workings of statistics/opinion research and just are taking for granted what Yougov provides for free.

Got it in one!

An ideal 'sample' would be 100% of the population being studied (and unbiased questions) but that is rarely possible. So studies have to use as small a sample as they can while still giving the same results as if the whole population were to be given the questions.

It is not just sample size, you also have to look at the composition of your sample. My fave was a study that in particular would be looking at low income people. So what did they originally choose as their *only* way to respond to the survey? Can we say the internet? This was a decade ago so I am sure a lot of people went without a network connection if it was that or heat.

Ok, you have your sample so now what? Your exact questions (and how they are asked) of course! Are you asking who will win an election or who will lose? Are you talking about issues or people? Do your questions favor one side over another? and so on ...

This is why you do not see the same exact number coming out of polls on the same question. There are way too many ways to skew results, even if unintentionally. Which is why you need to look at multiple polls to come up with an answer. Then you get into what weight you give each poll but for that you can start with how accurate the organization has been in past years.

 

[Innula Zenovka]

This is why I distrust them and prefer studies being done by proper, established opinion research companies over theirs any time.

That's the theory but, in practice, are YouGov's polling results notably out of line with those obtained by polling companies using more traditional methods? When they are, and when the results can be tested against reality (as in elections), are YouGov's results generally more or less accurate than those of their competitors?

I ask because I don't know, though my impression is that they're generally in line with their competitors, at least on UK political issues (you can compare the results from all UK polling companies at UK Polling Report).

 

[Porsupah Ree]

I can only echo their lament.

Andy UKpol 1/3

What I'd really like for Christmas is a Prime Minister who has a spine. A Prime Minister who will turn and say to her fellow politicians and her fellow uk citizens,
"As Leader of this country I’m going to stop this Brexit nonsense.
There have been no studies or reports detailing a plus side to any of this. I appreciate that a majority of you voted for it, but that was 2yrs ago when we only had the word of swivel-eyed loons like Johnson, Gove and Farage to "help" make our decision.

Andy UKpol 2/3

Their fantasies have been proven to be worthless. It was a shame that I nor anyone else on the Remain side didn't do more than sit around complacently underestimating pretty much everything.
I've now spent a lot of time and effort negotiating with the EU, and no-one is happy with deal. So maybe the concept of brexit is inherently flawed. Anything that causes the NHS and other organisations to start stock-piling supplies isn't a good thing.

Andy UKpol 3/3

Anything that prompts the government to have several thousand troops on standby, isn't a good thing.
So I've decided to call a stop to it. Those who really wanted it will be dead in a couple of years. Those who didn't want it, the young, would've suffered the most. So I am going to look beyond short-term political gain and look towards long-term gain for all the people of this country, the people of Europe, and cancel Article 50."

That, and an unlimited supply of Port and Stilton.

 

[Kara Spengler]

If you needed any more of a sign that brexit probably is not a good thing ....

Donnie pointed to brexit as a wonderful thing during his campaign and pretty much fellated farange on-stage. Meanwhile his supporters, most of whom did not know who anyone was outside of the US or about what other countries were doing, were not sure how they were supposed to react at first.

 

[Porsupah Ree]

It's almost as if even the Times is fed up with it all too..

 

[Luisa Land]

Quite simple: their own industry, opinion research companies.

Opinion research is based on statistics; in order to get a representative view - meaning that the views from the sample you have choosen can be applied to the rest of the population - you have got to choose the sample carefully, not leaving out any kind of age/social stratum/nationality. The holy grail you want to achieve is a so called random sample. If you don't know what that such a sample is: pulling out 10 colored balls of a bucket full of 100, and predicting the colors of the rest in it with best accuracy possible.

Working correctly to get a proper sample is hard; if you fail to draw a proper sample you got what is being called a statistical bias, meaning that the statements you are meaning cannot applied to the population. A typical bias could be leaving out certain social stratums, not having a real profile in your sample of the population, and so on.

The classic example on what a bias can cause happened in the 1930s in America, the polling desaster of 1936:

the popular magazine Literary Digest asked over 10 million readers about who they are going to vote for presidency, Landon or Roosevelt, and received 2,4 million answers. Literary Digest predicted a vote of 57% for Landon and 43% for Roosevelt. So many did believe this would come true.

There was another newcomer to the industry back then, Gallup. Gallup only interviewed around 50.000 people, and predicted the real outcome of 62% for Roosevelt and 38% for Landon. What happened? Quite simple: selection and non response bias. Literary Digest sent questions to people who had phone numbers back then, and asked them to return their answers. So they asked only the middle class, but entirely left out the lower social stratums, and many people of the 10 million they wrote to just didn't answer. Gallup in return manufactured his sample far more carefully, didn't leave out the lower income classes/poor and therefore was able to get far more accurate results using much less people.

This thing is the classical example of bias, how it ruins representativity and lectured to every student of statistics around the globe in their first course.

Coming back to Yougov and their main error: they are doing their public surveys online only. So this means you have got to sign up on their web site, answer polls and earn money. They are not using any other technic which is being used by classical opionion research institutions, like phone calls, asking people directly on the street, and so on - just the internet. So they always potentially do suffer a big, selectional bias by default regardless the question they do research.

By using just the internet with this approach you do have a really heavy bias, because you are unable to reach out to all the different social stratums you need to be able to hear to get a proper sample, and therefore proper results. This is why I distrust Yougov, because this simply cannot work for many questions if not all the people you need to ask are on the internet/internet affine. Many old people for example just won't care about such stuff in the internet by my experience, but you would need to ask them directly. They are therefore weighing their samples, but this is always another big, big source for errors, because at the end of the day it is always better to have a proper random sample than to recalculate your biased one based mostly on your experience. And this possible bias is rooted in the way Yougov works, it won't go away - ever, therefore weighing their samples is for them a must. This is why I distrust them and prefer studies being done by proper, established opinion research companies over theirs any time.

Selection Bias in UK Polling (Part 2): Internet Polling
https://www.displayr.com/what-is-selection-bias/
Why opinion pollsters failed to predict overall majority for David Cameron

The only real reasons why companies like Yougov, Civey and others are having customers is because they are really cheap compared to their much older competitors; if I would ever need to do serious market research, I would never trust anything Yougov/Civey and so on might come up with.

And most of the press articles where Yougov&Co. are being praised are being rewritten press notices by Yougov, where the authors don't even have the slighest grasp about the fundamental workings of statistics/opinion research and just are taking for granted what Yougov provides for free.

I still dont understand your argument:

You say YouGov uses online-research for their surveys and this is not as good as other methods (phone-calls or random-questions in the street..)
reading yourr post one could think, the problem is, who finds the Questionnaire, who knows that there is a survey, Some ppl might accidently find it. But only very few. Sure this would be far from representativ.
But this online research is done with a Panel, A Panel is a group of people of say we 10.000 or similar in this case so. peeople which are registred with the Opinion Pollster, YoGov or other similiar institutions are asking repeately these people over a long time
So they have a lot of data already of these persons and can exactly do what you say is the difficulty: the can build a sample of people according to criterions like age, occupation educational achievment, voter of labour, UkIP etc , undecided voter, living in a small, big city, countryside, other factors too they might think they are important for the subject ...it still stays difficult, but you can look or try to check, which factor might be important...
And I am realy sure, they prepare such interviews for example with checking questionaire with people before they start the survey in Pre-tests

the other methods of questioning makes it much more difficult if not impossible to build such a stratified sample
asking people in the streets you will have to ask many many people in many different parts of the country to get something what could be representativ., because in this way, you lack of information to pick up people in the streets is not building a sample at least if you will make statements about for example who will vote for the Tories.
besides you cant get all the information about the person, a panel provides, if you ask only once. Most people will not accept to be asked forhours.
and a Panel has the big advantage to recognize much easier developments , movements, : for example where do disapponted labour-voters tend to go: to which other parties or will they not vote next time etc.

What I read that more and more opinion-poll-institutions resign from research via phone-calls, cause of the poor response, More and more ppl dont accept to be called by phone

So as more as I think about it, as more I see the advantages of this method : a panel onlime-research..

I remember I once have been part of a panel, where the subject was to give opinions, judgements concerning movies, A movie-distributor liked to find out reactions of us, t0 bring the movie better to market.
And the project-director told us after some time, he was astonished how good that panel forecasted reactions of the movie-interested audience and at last the commercial success
Certainly not because we all were so great. I think just the method offers some chances, of course it has limitations..and always it needs researchers who know about the limitations and the chances

 

[Porsupah Ree]

Fuck.

Jeremy Corbyn has defiantly restated Labour’s policy of leading Britain out of the European Union with a refashioned Brexit deal, shrugging off intense pressure from Labour MPs and activists for the party to throw its weight behind a second referendum.

The Labour leader insisted that even if his party won a snap general election in the new year, he would seek to go to Brussels and try to secure a better deal – if possible, in time to allow Brexit to go ahead on 29 March.

Corbyn: Brexit would go ahead even if Labour won snap election

 

[Ashiri]

Fuck.

"Secure a better deal". WTF is wrong with these idiots?