Had the Entrance Exam disappeared by 1992? I suspect it had.
The Cambridge Sixth Term Entrance Paper (known to generations of students as STEP) is still used for Maths, usually on top of an A* or two at A-level.
Correct me if I'm wrong, but didn't they go (to be replaced just by interviews) and then come back?
That was my impression too. Cetainly Oxbridge has brought back preselection tests prior to holding interviews . That does not sound so very different from the earlier system in which Entrance Exam performance determined whether a candidate was invited for interview. What was different in those earlier years was that sucess at the Entrance Exam and the Interview meant that only 2 E grade passes at A level were needed to gain entry. If a candidate had already sat A levels,any offer given was unconditional.
The idea that a poll has an error of +/- 3% has been invalid for many years now. It hasn't been used in US polls recently and the Sturgis report recommended[1] that it no longer be used. The BPC agreed and issued a new rule[2] on 2018-05-01 saying that it should be replaced by this statement.
"All polls are subject to a wide range of potential sources of error. On the basis of the historical record of the polls at recent general elections, there is a 9 in 10 chance that the true value of a party’s support lies within 4 points of the estimates provided by this poll, and a 2 in 3 chance that they lie within 2 points."
Taking a more simple view, US way of life has become far too liberal, hence nearly 100% of people are on drugs, nearly 100% of under thirties have venereal disease. It needs conservatism to turn the country around.
To read geography at Exeter I was made an offer of an A in geography and 300 UCAS points. At the time the A in geography would be worth 120 meaning I needed 180 more. You got 60 for an A at AS Level so I could have ditched my other subjects and just finished my geography A Level and ditched the rest.
On topic: "So which pollsters’ numbers do you prefer?"
You shouldn't prefer any. You should take a moving average of them all. That will give a much more accurate picture.
The biggest source of variation is random error, typically +/- 3% in any individual poll.
Clearly incorrect. A moving average of all polls need not give a much more accurate picture.
If pollsters X has no systematic bias and pollster Y has, then averaging the polls of X and Y makes a more inaccurate result than using X alone.
In general, as the pollsters use different corrections (e.g., for turnout, for weighting, for treatment of undecideds), then averaging them all to produce a poll of polls is Frankenstein nonsense, statistically speaking.
You are simply hoping that the pollsters are all wrong in different ways and the errors cancel out.
The idea that a poll has an error of +/- 3% has been invalid for many years now. It hasn't been used in US polls recently and the Sturgis report recommended[1] that it no longer be used. The BPC agreed and issued a new rule[2] on 2018-05-01 saying that it should be replaced by this statement.
"All polls are subject to a wide range of potential sources of error. On the basis of the historical record of the polls at recent general elections, there is a 9 in 10 chance that the true value of a party’s support lies within 4 points of the estimates provided by this poll, and a 2 in 3 chance that they lie within 2 points."
Had the Entrance Exam disappeared by 1992? I suspect it had.
The Cambridge Sixth Term Entrance Paper (known to generations of students as STEP) is still used for Maths, usually on top of an A* or two at A-level.
Correct me if I'm wrong, but didn't they go (to be replaced just by interviews) and then come back?
I’m fairly certain the Maths ones never went away. I’ve taught a lot of students who had offers like A*A*AA STEP 1.
They were less than impressed when I told them what my offers were back in the day: two Cs from Imperial being the highest...
My Med School offer was BBC, though I got AAAB, similar to my peers. It gave a margin of safety when facing finals.
May I ask when that was? Are you referring to pre-late 1980s A levels? The assessment system changed radically thereafter - and grade inflation became rampant!
Remind me, when did poll cease to be quoted with the +/- 3% error margin?
If you chuck that back in, the polls are very consistent.
Indeed. People must be exceptionally bored to even bother to comment on these polls.
As Tim would say: MOE, ignore.
Has been thus for months.
Indeed, I don't think there has been a significant polling change in over a year. Its a stalemate. I cannot recall a longer period of stasis.
I agree
Yep, and I don’t think things will move much until after Brexit.
Will be interesting to see where the Remainers go after that...
Or indeed diehard Leavers given it will largely be BINO after next March with the UK committed to staying in the Customs Union until a technical solution is agreed to the resolve the Irish border
On topic: "So which pollsters’ numbers do you prefer?"
You shouldn't prefer any. You should take a moving average of them all. That will give a much more accurate picture.
The biggest source of variation is random error, typically +/- 3% in any individual poll.
Clearly incorrect. A moving average of all polls need not give a much more accurate picture.
If pollsters X has no systematic bias and pollster Y has, then averaging the polls of X and Y makes a more inaccurate result than using X alone.
In general, as the pollsters use different corrections (e.g., for turnout, for weighting, for treatment of undecideds), then averaging them all to produce a poll of polls is Frankenstein nonsense, statistically speaking.
You are simply hoping that the pollsters are all wrong in different ways and the errors cancel out.
But, there is no reason to believe this.
I also don’t agree that the biggest source of error is random. There is plenty of evidence that the biggest source of error is systematic.
You are statistically confused and using unsupported conjecture.
We are talking about best estimates.
If you don't know whether pollster X or Y has systematic bias, then the average of the two is the best estimate. Obviously if you do know which has no bias, that's the one to follow. But you don't know.
All pollsters are shooting at a target of being the most accurate. As you say they use different methodologies. But averaging them, in the absence of knowledge of which is the most accurate methodology, produces the best estimate. To say this is "Frankenstein nonsense" is to use emotive language to cover up a lack of understanding.
You say "There is plenty of evidence that the biggest source of error is systematic." Can you point to that evidence? Do you agree that the random error is about +/- 3% about 95% of the time? 5% of the time you get outliers outside that range. Do you think that the systematic error of pollsters is generally greater than 3%? If so, why?
Economics is an interesting example of the snobbishness in our university system. It is accepted that LSE is the top economics school in the land. But the second is Bath, which is not a member of the Russell Group and therefore cannot recruit beyond a certain level. That in itself underlines the stupidity of that particular reform.
As an aside, I have a school friend who went to Bath to do economics. I think she did OK for herself in the end. She's now a Professor at some small university in the USA, called Stamford Business School I think. Some of you may have heard of her as well - her name is Alison Goldsworthy and she's a former senior Liberal Democrat student activist, one of the women who had the misfortune to be involved in the Rennard case.
The idea that a poll has an error of +/- 3% has been invalid for many years now. It hasn't been used in US polls recently and the Sturgis report recommended[1] that it no longer be used. The BPC agreed and issued a new rule[2] on 2018-05-01 saying that it should be replaced by this statement.
"All polls are subject to a wide range of potential sources of error. On the basis of the historical record of the polls at recent general elections, there is a 9 in 10 chance that the true value of a party’s support lies within 4 points of the estimates provided by this poll, and a 2 in 3 chance that they lie within 2 points."
Off topic somewhat I watched a bit of Dateline on News 24 - all 4 panellists anti May with Yasmin Ali and that absurd Shabba representing British journalists with a PPB for the Labour party. Extraordinarily biased.
Remind me, when did poll cease to be quoted with the +/- 3% error margin?
If you chuck that back in, the polls are very consistent.
Indeed. People must be exceptionally bored to even bother to comment on these polls.
As Tim would say: MOE, ignore.
Has been thus for months.
Indeed, I don't think there has been a significant polling change in over a year. Its a stalemate. I cannot recall a longer period of stasis.
I agree
Yep, and I don’t think things will move much until after Brexit.
Will be interesting to see where the Remainers go after that...
Or indeed diehard Leavers given it will largely be BINO after next March with the UK committed to staying in the Customs Union until a technical solution is agreed to the resolve the Irish border
The transition proposed is a no change transition as far as SM/CU is concerned. I still don’t think you’ve understood that.
How - how could people be 'cheated out of Brexit' by another referendum? It'd need the people to vote to Remain after all, which they won't do if they want Brexit after all. Were the people 'cheated out of' a Conservative majority last year?
Because we haven’t implemented the result of the last one yet...
If you don't know whether pollster X or Y has systematic bias, then the average of the two is the best estimate. Obviously if you do know which has no bias, that's the one to follow. But you don't know.
You have two clocks. One of them says 11:55. Another says 12:05. What's the right time?
The average of two polls is not necessarily more accurate than one of the polls. In the EU referendum the online polls were more accurate than the the telephone polls and the average of them was not the closest to the truth. When the experts looked at this (eg [3]) they were all over the place.
The average is closer to the truth in the case where all the samples are done using the same method at the same time from the same sample frame, providing the method is random and the frame representative. In that case the Central Limit Theorem kicks in and the samples vary normally around the truth, so a poll-of-polls has merit.
But IRL the samples are nonrandom, done using different methods, the companies use different weights and different sample frames, and VI changes over time. The CRT doesn't apply here and there's no theoretical justification for assuming a poll-of-polls is better.
Off topic somewhat I watched a bit of Dateline on News 24 - all 4 panellists anti May with Yasmin Ali and that absurd Shabba representing British journalists with a PPB for the Labour party. Extraordinarily biased.
I watched that in astonishment. It is unacceptable but is constant with Dateline
How - how could people be 'cheated out of Brexit' by another referendum? It'd need the people to vote to Remain after all, which they won't do if they want Brexit after all. Were the people 'cheated out of' a Conservative majority last year?
Because we haven’t implemented the result of the last one yet...
The last one was too vague
That’s not something that can be decided after the vote went against your wishes.
Remind me, when did poll cease to be quoted with the +/- 3% error margin?
If you chuck that back in, the polls are very consistent.
Indeed. People must be exceptionally bored to even bother to comment on these polls.
As Tim would say: MOE, ignore.
Has been thus for months.
Indeed, I don't think there has been a significant polling change in over a year. Its a stalemate. I cannot recall a longer period of stasis.
I agree
Yep, and I don’t think things will move much until after Brexit.
Will be interesting to see where the Remainers go after that...
Or indeed diehard Leavers given it will largely be BINO after next March with the UK committed to staying in the Customs Union until a technical solution is agreed to the resolve the Irish border
The transition proposed is a no change transition as far as SM/CU is concerned. I still don’t think you’ve understood that.
I think HYUFD is pointing out that we will stay in the Customs Union beyond the transition period until a technical solution is agreed to the resolve the Irish border. That could be for a long time as it has to be agreed by the EU.
The Senate has therefore voted to confirm Kavanaugh to be appointed as a US Supreme Court Justice
Which 2 didn't vote?
Presumably the guy at his daughter's wedding. I wonder is someone was decent enough to give him the ability to do so by undertaking not to vote either.
Had the Entrance Exam disappeared by 1992? I suspect it had.
The Cambridge Sixth Term Entrance Paper (known to generations of students as STEP) is still used for Maths, usually on top of an A* or two at A-level.
Correct me if I'm wrong, but didn't they go (to be replaced just by interviews) and then come back?
I’m fairly certain the Maths ones never went away. I’ve taught a lot of students who had offers like A*A*AA STEP 1.
They were less than impressed when I told them what my offers were back in the day: two Cs from Imperial being the highest...
My Med School offer was BBC, though I got AAAB, similar to my peers. It gave a margin of safety when facing finals.
You need a bit more than BBC now...
I know, I amon our admissions panel. Generally AAA, but we let in AAB as long as Chemistry is an A.
My brother got into LSE with A levels in Maths, Physics and Economics, but it was 1981!
It's the LSE, Dr. They haven't changed since the 1890s never mind the 1980s!
Its much more right wing nowadays. When my brother was there the Union coverd the political spectrum from Trotskyism to Communism.
Find it ironic people still think LSE is left wing. When I was there 30 years ago it provided much of the intellectual heft behind Thatcherism . Was where Hayek was after all .
Remind me, when did poll cease to be quoted with the +/- 3% error margin?
If you chuck that back in, the polls are very consistent.
Indeed. People must be exceptionally bored to even bother to comment on these polls.
As Tim would say: MOE, ignore.
Has been thus for months.
Indeed, I don't think there has been a significant polling change in over a year. Its a stalemate. I cannot recall a longer period of stasis.
I agree
Yep, and I don’t think things will move much until after Brexit.
Will be interesting to see where the Remainers go after that...
Or indeed diehard Leavers given it will largely be BINO after next March with the UK committed to staying in the Customs Union until a technical solution is agreed to the resolve the Irish border
The transition proposed is a no change transition as far as SM/CU is concerned. I still don’t think you’ve understood that.
It is until December 2020 but even after December 2020 now May is set to confirm the backstop for Northern Ireland is the whole UK stays in the Customs Union no FTA can be agreed that sees the UK leave the Customs Union unless a technical solution has been agreed between the UK and EU on avoiding a hard border in Ireland
US Senate voting is just crazy. I am not sure what the 30 hour delay was all about. It didn't change a thing. Allowing senators to change their votes is crazy.
The Senate has therefore voted to confirm Kavanaugh to be appointed as a US Supreme Court Justice
Which 2 didn't vote?
Presumably the guy at his daughter's wedding. I wonder is someone was decent enough to give him the ability to do so by undertaking not to vote either.
Taking a more simple view, US way of life has become far too liberal, hence nearly 100% of people are on drugs, nearly 100% of under thirties have venereal disease. It needs conservatism to turn the country around.
Could I have a source for those figures please?
100% imaginary, I believe.
Not only imaginary, but wildly inaccurate. All the evidence is that young people today are drinking less and generally behaving more healthily than people of the previous generation. (I.e. me.)
STDs are increasing (although the group where they are rising fastest is not the young, but the old), but that's not because people are having more sexual partners, but because we've all stopped worrying about AIDS.
The idea that a poll has an error of +/- 3% has been invalid for many years now. It hasn't been used in US polls recently and the Sturgis report recommended[1] that it no longer be used. The BPC agreed and issued a new rule[2] on 2018-05-01 saying that it should be replaced by this statement.
"All polls are subject to a wide range of potential sources of error. On the basis of the historical record of the polls at recent general elections, there is a 9 in 10 chance that the true value of a party’s support lies within 4 points of the estimates provided by this poll, and a 2 in 3 chance that they lie within 2 points."
OK That simply means the standard deviation is 2% not 1.5%. Random variation is even greater than I was suggesting.
It's a bit more fundamental than that. From the Sturgis report:
"The inquiry was also tasked with assessing whether uncertainty in poll estimates is adequately communicated to stakeholders and the public. Our conclusion is that it is not. Media commentators generally use a rule of thumb that point estimates of party shares have a margin of error of + or - three percentage points. This is (or appears to be) based on the calculation of a 95% confidence interval for a simple random sample of 1000 respondents for a party with 50% of the vote. But polls are not simple random samples, parties are typically far from 50% of the vote, and when the full sample is 1000 the number of respondents used for estimating party shares is much less than this number. For these reasons, the true margin of error will usually be different from + or – 3 per cent, sometimes larger and sometimes smaller, depending on the specifics of the research design. "
Much of the terms used in polls ("margin of error", "standard deviation", "confidence interval") have their roots in statistical theory. But the compromises required to get a poll out quickly and cheaply (nonrandom sampling, nonrepresentative models, turnout adjustments) push the theory beyond breaking point. This is one on the reasons that recommendation was based on past results instead of a theoretical formula: the theory doesn't apply any more and hasn't done for many years now.
The idea that a poll has an error of +/- 3% has been invalid for many years now. It hasn't been used in US polls recently and the Sturgis report recommended[1] that it no longer be used. The BPC agreed and issued a new rule[2] on 2018-05-01 saying that it should be replaced by this statement.
"All polls are subject to a wide range of potential sources of error. On the basis of the historical record of the polls at recent general elections, there is a 9 in 10 chance that the true value of a party’s support lies within 4 points of the estimates provided by this poll, and a 2 in 3 chance that they lie within 2 points."
US Senate voting is just crazy. I am not sure what the 30 hour delay was all about. It didn't change a thing. Allowing senators to change their votes is crazy.
It is no way to run anything.
I predict that the US abortion laws won't change. All that will happen is that one or two states like Mississippi will try to ban it but it'll get held up in legal wrangles for many years and nothing much will happen.
I wouldn't normally bother linking to a Guido-snipe, and his style is indicative of many of the problems I am about to criticise, but this one was well-executed. Still, it leaves me in some despair - what have us early 21st-century West-dwellers done, to have built for ourselves a high-tech communication mechanism for the great affairs of the day, but have somehow done so in the form of Twitter, and to have created for ourselves (and it must be us, the public, who are somehow collectively paying for their continued sustenance) an elite commentariat such as this?
If you don't know whether pollster X or Y has systematic bias, then the average of the two is the best estimate. Obviously if you do know which has no bias, that's the one to follow. But you don't know.
You have two clocks. One of them says 11:55. Another says 12:05. What's the right time?
The average of two polls is not necessarily more accurate than one of the polls. In the EU referendum the online polls were more accurate than the the telephone polls and the average of them was not the closest to the truth. When the experts looked at this (eg [3]) they were all over the place.
The average is closer to the truth in the case where all the samples are done using the same method at the same time from the same sample frame, providing the method is random and the frame representative. In that case the Central Limit Theorem kicks in and the samples vary normally around the truth, so a poll-of-polls has merit.
But IRL the samples are nonrandom, done using different methods, the companies use different weights and different sample frames, and VI changes over time. The CRT doesn't apply here and there's no theoretical justification for assuming a poll-of-polls is better.
Your clock example is spurious. We are trying to ascertain the best estimate. It won't be correct. It will be just be the best estimate for whatever purpose you have in mind - like what is the likely composition of the next government.
I know the average of two polls is not necessarily more accurate than one of the polls. We all know that. But we don't know which poll! That's the point.
I know all about the Central Limit Theorem and that the theoretical conditions don't apply with mixed polling methodologies. But I'm a practical mathematician not a pure one, trying to get a feel of what is the likely shape of the next government based on current polls, rather producing a water tight statistical paper than will pass peer review.
US Senate voting is just crazy. I am not sure what the 30 hour delay was all about. It didn't change a thing. Allowing senators to change their votes is crazy.
It is no way to run anything.
Unless he exhibits considerably more mental balance and judgement than he did in his confirmation hearings, I predict this is anything but over. It could easily be just beginning.
I wouldn't normally bother linking to a Guido-snipe, and his style is indicative of many of the problems I am about to criticise, but this one was well-executed. Still, it leaves me in some despair - what have us early 21st-century West-dwellers done, to have built for ourselves a high-tech communication mechanism for the great affairs of the day, but have somehow done so in the form of Twitter, and to have created for ourselves (and it must be us, the public, who are somehow collectively paying for their continued sustenance) an elite commentariat such as this?
That first tweet was cleverly worded by Maguire. He didn't actually say it was because of Brexit, just that it would be bad for Brexit Britain, though it does imply that losing Unilever wouldn't be so bad if we weren't leaving the EU.
And I speak as someone who thinks Roe v Wade is an abomination.
You think it is an abomination that a woman has a right to safe medical treatment?
It's a view, I suppose ...
.
Taking a more simple view, US way of life has become far too liberal, hence nearly 100% of people are on drugs, nearly 100% of under thirties have venereal disease. It needs conservatism to turn the country around.
US Senate voting is just crazy. I am not sure what the 30 hour delay was all about. It didn't change a thing. Allowing senators to change their votes is crazy.
It is no way to run anything.
Unless he exhibits considerably more mental balance and judgement than he did in his confirmation hearings, I predict this is anything but over. It could easily be just beginning.
I would have thought that with the most intense and stressful period of his life, for the most part, now behind him, that the chances would be much higher that he would indeed be able to exhibit more balance and judgement, sadly for those hoping for something that will keep it going.
The idea that a poll has an error of +/- 3% has been invalid for many years now. It hasn't been used in US polls recently and the Sturgis report recommended[1] that it no longer be used. The BPC agreed and issued a new rule[2] on 2018-05-01 saying that it should be replaced by this statement.
"All polls are subject to a wide range of potential sources of error. On the basis of the historical record of the polls at recent general elections, there is a 9 in 10 chance that the true value of a party’s support lies within 4 points of the estimates provided by this poll, and a 2 in 3 chance that they lie within 2 points."
OK That simply means the standard deviation is 2% not 1.5%. Random variation is even greater than I was suggesting.
It's a bit more fundamental than that. From the Sturgis report:
"The inquiry was also tasked with assessing whether uncertainty in poll estimates is adequately communicated to stakeholders and the public. Our conclusion is that it is not. Media commentators generally use a rule of thumb that point estimates of party shares have a margin of error of + or - three percentage points. This is (or appears to be) based on the calculation of a 95% confidence interval for a simple random sample of 1000 respondents for a party with 50% of the vote. But polls are not simple random samples, parties are typically far from 50% of the vote, and when the full sample is 1000 the number of respondents used for estimating party shares is much less than this number. For these reasons, the true margin of error will usually be different from + or – 3 per cent, sometimes larger and sometimes smaller, depending on the specifics of the research design. "
Much of the terms used in polls ("margin of error", "standard deviation", "confidence interval") have their roots in statistical theory. But the compromises required to get a poll out quickly and cheaply (nonrandom sampling, nonrepresentative models, turnout adjustments) push the theory beyond breaking point. This is one on the reasons that recommendation was based on past results instead of a theoretical formula: the theory doesn't apply any more and hasn't done for many years now.
Quite. Use past results rather than statistical theory. My point remains.
And I speak as someone who thinks Roe v Wade is an abomination.
You think it is an abomination that a woman has a right to safe medical treatment?
It's a view, I suppose ...
.
Taking a more simple view, US way of life has become far too liberal, hence nearly 100% of people are on drugs, nearly 100% of under thirties have venereal disease. It needs conservatism to turn the country around.
Your clock example is spurious. We are trying to ascertain the best estimate. It won't be correct. It will be just be the best estimate for whatever purpose you have in mind - like what is the likely composition of the next government.
I know the average of two polls is not necessarily more accurate than one of the polls. We all know that. But we don't know which poll! That's the point.
I know about the "don't know which poll!" problem. But the solution is not to take an average, it's to say "I don't know". This is why I try not to track the poll-of-polls. I know your approach is to take a moving average of polls and track that, but my approach is to track polls from each company.
Taking a more simple view, US way of life has become far too liberal, hence nearly 100% of people are on drugs, nearly 100% of under thirties have venereal disease. It needs conservatism to turn the country around.
Could I have a source for those figures please?
100% imaginary, I believe.
Not only imaginary, but wildly inaccurate. All the evidence is that young people today are drinking less and generally behaving more healthily than people of the previous generation. (I.e. me.)
STDs are increasing (although the group where they are rising fastest is not the young, but the old), but that's not because people are having more sexual partners, but because we've all stopped worrying about AIDS.
As you state, the trend may not seem bad to you, but to me these are very high and very sad %s. If it’s about impacts, then surely such things can break relationships, families, wreck homes, break people? And that means in the bigger picture, problems for society with crime, state support, mental health.
For GB I suspect lower %s than US, but If it’s about trends, then surely trend in US and GB is in the wrong direction? Here in GB, police are saying to middle class drug users, you are fuelling crime epidemic, are they not?
Some dismiss my posts as unserious others claim I shouldn’t be called LordOfReason. But I think to ask is it liberal politicians or Conservative politicians who have the values to bring these %s down is a reasonable question to ask.
Polling isn't counting a sample of red and blue marbles in a sack.
The biggest fundamental problem to me is probably differential non response, closely followed by people outright lieing to the pollster with people changing their god damned minds at the ballot box being another issue.
Your clock example is spurious. We are trying to ascertain the best estimate. It won't be correct. It will be just be the best estimate for whatever purpose you have in mind - like what is the likely composition of the next government.
I know the average of two polls is not necessarily more accurate than one of the polls. We all know that. But we don't know which poll! That's the point.
I know about the "don't know which poll!" problem. But the solution is not to take an average, it's to say "I don't know". This is why I try not to track the poll-of-polls. I know your approach is to take a moving average of polls and track that, but my approach is to track polls from each company.
That's fine. I understand why you might choose to do that. But how do you take a view of most likely outcome? You end up tracking polls from each company but to what end?
Polling isn't counting a sample of red and blue marbles in a sack.
The biggest fundamental problem to me is probably differential non response, closely followed by people outright lieing to the pollster with people changing their god damned minds at the ballot box being another issue.
Least we don't have people generally lieing in the exit polls *Cough Netanyahu*...
Had the Entrance Exam disappeared by 1992? I suspect it had.
The Cambridge Sixth Term Entrance Paper (known to generations of students as STEP) is still used for Maths, usually on top of an A* or two at A-level.
Correct me if I'm wrong, but didn't they go (to be replaced just by interviews) and then come back?
I’m fairly certain the Maths ones never went away. I’ve taught a lot of students who had offers like A*A*AA STEP 1.
They were less than impressed when I told them what my offers were back in the day: two Cs from Imperial being the highest...
My Med School offer was BBC, though I got AAAB, similar to my peers. It gave a margin of safety when facing finals.
May I ask when that was? Are you referring to pre-late 1980s A levels? The assessment system changed radically thereafter - and grade inflation became rampant!
1983.
@Fysicsteacher is right about predicted grades. Private schools notoriously over egg them, some more than others. One of the most effective way my University has widened access fairly is by making post results offers. Basing on actual rather than predicted results has substantially increased the number of successful state school admissions. The extra year gives added maturity too.
Your clock example is spurious. We are trying to ascertain the best estimate. It won't be correct. It will be just be the best estimate for whatever purpose you have in mind - like what is the likely composition of the next government.
I know the average of two polls is not necessarily more accurate than one of the polls. We all know that. But we don't know which poll! That's the point.
I know about the "don't know which poll!" problem. But the solution is not to take an average, it's to say "I don't know". This is why I try not to track the poll-of-polls. I know your approach is to take a moving average of polls and track that, but my approach is to track polls from each company.
That's fine. I understand why you might choose to do that. But how do you take a view of most likely outcome? You end up tracking polls from each company but to what end?
Yes, one never knows which pollster is correct ! Yougov's 'map' seemed to work well for 2017, hopefully they'll do it again.
Off topic somewhat I watched a bit of Dateline on News 24 - all 4 panellists anti May with Yasmin Ali and that absurd Shabba representing British journalists with a PPB for the Labour party. Extraordinarily biased.
I watched that in astonishment. It is unacceptable but is constant with Dateline
I do not know how they get away with it
I chanced upon it too. Utterly disgraceful programme. And the BBC pump it out on their international network too where it will be viewed as impartial and factual. Anyone who thinks the BBC isn't a hotbed of left wing bias is sadly deluded.
Polling isn't counting a sample of red and blue marbles in a sack.
The biggest fundamental problem to me is probably differential non response, closely followed by people outright lieing to the pollster with people changing their god damned minds at the ballot box being another issue.
This is all true. Have you ever been out canvassing? Nevertheless if you want to get a feel of the current likely outcome of a general election now, the polls are your best tool. Scary.
I wouldn't normally bother linking to a Guido-snipe, and his style is indicative of many of the problems I am about to criticise, but this one was well-executed. Still, it leaves me in some despair - what have us early 21st-century West-dwellers done, to have built for ourselves a high-tech communication mechanism for the great affairs of the day, but have somehow done so in the form of Twitter, and to have created for ourselves (and it must be us, the public, who are somehow collectively paying for their continued sustenance) an elite commentariat such as this?
That first tweet was cleverly worded by Maguire. He didn't actually say it was because of Brexit, just that it would be bad for Brexit Britain, though it does imply that losing Unilever wouldn't be so bad if we weren't leaving the EU.
Yes, the kind of cleverness that does a commentator no credit though. I did think Guido was being a bit unfair here, if you read the text carefully, then I thought - nah.
Guido's as bad at cherry-picking and repackaging facts to make a point as anyone else so it's an act of hypocrisy on his part when he calls people out on it. It's just frustrating that the entire commentariat seem to be this bad and twitter mayhaps has led to further degeneration.
If professional political commentary roles were allocated on the grounds of quality and insight, David Herdson would be one of this country's foremost columnists and talking heads.
Taking a more simple view, US way of life has become far too liberal, hence nearly 100% of people are on drugs, nearly 100% of under thirties have venereal disease. It needs conservatism to turn the country around.
Could I have a source for those figures please?
100% imaginary, I believe.
Not only imaginary, but wildly inaccurate. All the evidence is that young people today are drinking less and generally behaving more healthily than people of the previous generation. (I.e. me.)
STDs are increasing (although the group where they are rising fastest is not the young, but the old), but that's not because people are having more sexual partners, but because we've all stopped worrying about AIDS.
As you state, the trend may not seem bad to you, but to me these are very high and very sad %s. If it’s about impacts, then surely such things can break relationships, families, wreck homes, break people? And that means in the bigger picture, problems for society with crime, state support, mental health.
For GB I suspect lower %s than US, but If it’s about trends, then surely trend in US and GB is in the wrong direction? Here in GB, police are saying to middle class drug users, you are fuelling crime epidemic, are they not?
Some dismiss my posts as unserious others claim I shouldn’t be called LordOfReason. But I think to ask is it liberal politicians or Conservative politicians who have the values to bring these %s down is a reasonable question to ask.
Your clock example is spurious. We are trying to ascertain the best estimate. It won't be correct. It will be just be the best estimate for whatever purpose you have in mind - like what is the likely composition of the next government.
I know the average of two polls is not necessarily more accurate than one of the polls. We all know that. But we don't know which poll! That's the point.
I know about the "don't know which poll!" problem. But the solution is not to take an average, it's to say "I don't know". This is why I try not to track the poll-of-polls. I know your approach is to take a moving average of polls and track that, but my approach is to track polls from each company.
That's fine. I understand why you might choose to do that. But how do you take a view of most likely outcome? You end up tracking polls from each company but to what end?
Identifying the trend and the ballpark. If one cannot get a point estimate, one can at least say "well, it's somewhere in that area".
OK, even that statement is wrong (as the EU ref proved) but it's a bit more reliable.
Your clock example is spurious. We are trying to ascertain the best estimate. It won't be correct. It will be just be the best estimate for whatever purpose you have in mind - like what is the likely composition of the next government.
I know the average of two polls is not necessarily more accurate than one of the polls. We all know that. But we don't know which poll! That's the point.
I know about the "don't know which poll!" problem. But the solution is not to take an average, it's to say "I don't know". This is why I try not to track the poll-of-polls. I know your approach is to take a moving average of polls and track that, but my approach is to track polls from each company.
That's fine. I understand why you might choose to do that. But how do you take a view of most likely outcome? You end up tracking polls from each company but to what end?
Yes, one never knows which pollster is correct ! Yougov's 'map' seemed to work well for 2017, hopefully they'll do it again.
As I explained downthread, YouGov do seem to have a small bias. If I have time, I calculate the bias of each of the main pollsters.
How - how could people be 'cheated out of Brexit' by another referendum? It'd need the people to vote to Remain after all, which they won't do if they want Brexit after all. Were the people 'cheated out of' a Conservative majority last year?
Because we haven’t implemented the result of the last one yet...
Like we'd implemented the entire 2015 Tory manifesto? No way. Obviously, the 2017 GE did not respect the result of the 2015 GE and calling it was democratically illegitimate.
How - how could people be 'cheated out of Brexit' by another referendum? It'd need the people to vote to Remain after all, which they won't do if they want Brexit after all. Were the people 'cheated out of' a Conservative majority last year?
Because we haven’t implemented the result of the last one yet...
Like we'd implemented the entire 2015 Tory manifesto? No way. Obviously, the 2017 GE did not respect the result of the 2015 GE and calling it was democratically illegitimate.
Opportunistic, unwise and unnecessary certainly but how can holding an election be democratically illegitimate?
How - how could people be 'cheated out of Brexit' by another referendum? It'd need the people to vote to Remain after all, which they won't do if they want Brexit after all. Were the people 'cheated out of' a Conservative majority last year?
Because we haven’t implemented the result of the last one yet...
Like we'd implemented the entire 2015 Tory manifesto? No way. Obviously, the 2017 GE did not respect the result of the 2015 GE and calling it was democratically illegitimate.
The Tories had formed a government after the 2015 GE. The fact they decided to then go back to the polls earlier than expected is neither here nor there, the 2015 results had been honoured.
The 2016 results haven't been yet. We've not left yet. If we have another referendum after we leave but before all Vote Leave's promises have been honoured then that would be comparable.
Your clock example is spurious. We are trying to ascertain the best estimate. It won't be correct. It will be just be the best estimate for whatever purpose you have in mind - like what is the likely composition of the next government.
I know the average of two polls is not necessarily more accurate than one of the polls. We all know that. But we don't know which poll! That's the point.
I know about the "don't know which poll!" problem. But the solution is not to take an average, it's to say "I don't know". This is why I try not to track the poll-of-polls. I know your approach is to take a moving average of polls and track that, but my approach is to track polls from each company.
That's fine. I understand why you might choose to do that. But how do you take a view of most likely outcome? You end up tracking polls from each company but to what end?
Identifying the trend and the ballpark. If one cannot get a point estimate, one can at least say "well, it's somewhere in that area".
OK, even that statement is wrong (as the EU ref proved) but it's a bit more reliable.
Yes -point estimates are misleading. Ballpark estimates are more useful and less misleading. It's somewhere between 11:55 and 12:05.
How - how could people be 'cheated out of Brexit' by another referendum? It'd need the people to vote to Remain after all, which they won't do if they want Brexit after all. Were the people 'cheated out of' a Conservative majority last year?
Because we haven’t implemented the result of the last one yet...
Like we'd implemented the entire 2015 Tory manifesto? No way. Obviously, the 2017 GE did not respect the result of the 2015 GE and calling it was democratically illegitimate.
They became the government.
And implemented a popular part of their manifesto - an in/out referendum.
How - how could people be 'cheated out of Brexit' by another referendum? It'd need the people to vote to Remain after all, which they won't do if they want Brexit after all. Were the people 'cheated out of' a Conservative majority last year?
Because we haven’t implemented the result of the last one yet...
Like we'd implemented the entire 2015 Tory manifesto? No way. Obviously, the 2017 GE did not respect the result of the 2015 GE and calling it was democratically illegitimate.
Opportunistic, unwise and unnecessary certainly but how can holding an election be democratically illegitimate?
No more than holding a referendum is or is not so. I agree. I was employing sarcasm.
As you state, the trend may not seem bad to you, but to me these are very high and very sad %s. If it’s about impacts, then surely such things can break relationships, families, wreck homes, break people? And that means in the bigger picture, problems for society with crime, state support, mental health.
For GB I suspect lower %s than US, but If it’s about trends, then surely trend in US and GB is in the wrong direction? Here in GB, police are saying to middle class drug users, you are fuelling crime epidemic, are they not?
Some dismiss my posts as unserious others claim I shouldn’t be called LordOfReason. But I think to ask is it liberal politicians or Conservative politicians who have the values to bring these %s down is a reasonable question to ask.
So, 25% is basically the same as 100%?
Wow.
At least your not discounting these figures. I’ll agree it’s clearly not (with caveat though all these stats likely somewhat higher in reality rather than lower, we’ll never know the true picture).
Do you agree these are massively high stats for drug problem and STD rates?
is it liberal politicians and judges or Conservative politicians and judges who have the values to bring these %s down would you say?
How - how could people be 'cheated out of Brexit' by another referendum? It'd need the people to vote to Remain after all, which they won't do if they want Brexit after all. Were the people 'cheated out of' a Conservative majority last year?
Because we haven’t implemented the result of the last one yet...
Like we'd implemented the entire 2015 Tory manifesto? No way. Obviously, the 2017 GE did not respect the result of the 2015 GE and calling it was democratically illegitimate.
Opportunistic, unwise and unnecessary certainly but how can holding an election be democratically illegitimate?
No more than holding a referendum is or is not so. I agree. I was employing sarcasm.
Holding a new referendum after we've exited the EU is entirely reasonable.
As was holding a new election after the last election's winners had formed a government.
Holding a referendum now would be like Corbyn winning the 2017 election and then protestors saying "lets have another election to check you really wanted Corbyn in power" before he entered Downing Street.
We voted to Leave, lets Leave and then we can talk about new votes.
I wouldn't normally bother linking to a Guido-snipe, and his style is indicative of many of the problems I am about to criticise, but this one was well-executed. Still, it leaves me in some despair - what have us early 21st-century West-dwellers done, to have built for ourselves a high-tech communication mechanism for the great affairs of the day, but have somehow done so in the form of Twitter, and to have created for ourselves (and it must be us, the public, who are somehow collectively paying for their continued sustenance) an elite commentariat such as this?
That first tweet was cleverly worded by Maguire. He didn't actually say it was because of Brexit, just that it would be bad for Brexit Britain, though it does imply that losing Unilever wouldn't be so bad if we weren't leaving the EU.
Yes, the kind of cleverness that does a commentator no credit though. I did think Guido was being a bit unfair here, if you read the text carefully, then I thought - nah.
Guido's as bad at cherry-picking and repackaging facts to make a point as anyone else so it's an act of hypocrisy on his part when he calls people out on it. It's just frustrating that the entire commentariat seem to be this bad and twitter mayhaps has led to further degeneration.
If professional political commentary roles were allocated on the grounds of quality and insight, David Herdson would be one of this country's foremost columnists and talking heads.
How - how could people be 'cheated out of Brexit' by another referendum? It'd need the people to vote to Remain after all, which they won't do if they want Brexit after all. Were the people 'cheated out of' a Conservative majority last year?
Because we haven’t implemented the result of the last one yet...
Like we'd implemented the entire 2015 Tory manifesto? No way. Obviously, the 2017 GE did not respect the result of the 2015 GE and calling it was democratically illegitimate.
The Tories had formed a government after the 2015 GE. The fact they decided to then go back to the polls earlier than expected is neither here nor there, the 2015 results had been honoured.
The 2016 results haven't been yet. We've not left yet. If we have another referendum after we leave but before all Vote Leave's promises have been honoured then that would be comparable.
The vote was for a narrow Tory majority to govern for five years according to the programme for government they'd put forwards. Both the narrow majority and the programme in question were inconvenient for the Government, so they sought to overturn that result opportunistically.
How - how could people be 'cheated out of Brexit' by another referendum? It'd need the people to vote to Remain after all, which they won't do if they want Brexit after all. Were the people 'cheated out of' a Conservative majority last year?
Because we haven’t implemented the result of the last one yet...
Like we'd implemented the entire 2015 Tory manifesto? No way. Obviously, the 2017 GE did not respect the result of the 2015 GE and calling it was democratically illegitimate.
The Tories had formed a government after the 2015 GE. The fact they decided to then go back to the polls earlier than expected is neither here nor there, the 2015 results had been honoured.
The 2016 results haven't been yet. We've not left yet. If we have another referendum after we leave but before all Vote Leave's promises have been honoured then that would be comparable.
The vote was for a narrow Tory majority to govern for five years according to the programme for government they'd put forwards. Both the narrow majority and the programme in question were inconvenient for the Government, so they sought to overturn that result opportunistically.
How - how could people be 'cheated out of Brexit' by another referendum? It'd need the people to vote to Remain after all, which they won't do if they want Brexit after all. Were the people 'cheated out of' a Conservative majority last year?
Because we haven’t implemented the result of the last one yet...
Like we'd implemented the entire 2015 Tory manifesto? No way. Obviously, the 2017 GE did not respect the result of the 2015 GE and calling it was democratically illegitimate.
The Tories had formed a government after the 2015 GE. The fact they decided to then go back to the polls earlier than expected is neither here nor there, the 2015 results had been honoured.
The 2016 results haven't been yet. We've not left yet. If we have another referendum after we leave but before all Vote Leave's promises have been honoured then that would be comparable.
The vote was for a narrow Tory majority to govern for five years according to the programme for government they'd put forwards. Both the narrow majority and the programme in question were inconvenient for the Government, so they sought to overturn that result opportunistically.
But they'd formed the government.
Once we've left we can have another vote. Even if its opportunistic. We haven't left yet so its not comparable.
How - how could people be 'cheated out of Brexit' by another referendum? It'd need the people to vote to Remain after all, which they won't do if they want Brexit after all. Were the people 'cheated out of' a Conservative majority last year?
Because we haven’t implemented the result of the last one yet...
Like we'd implemented the entire 2015 Tory manifesto? No way. Obviously, the 2017 GE did not respect the result of the 2015 GE and calling it was democratically illegitimate.
The Tories had formed a government after the 2015 GE. The fact they decided to then go back to the polls earlier than expected is neither here nor there, the 2015 results had been honoured.
The 2016 results haven't been yet. We've not left yet. If we have another referendum after we leave but before all Vote Leave's promises have been honoured then that would be comparable.
The vote was for a narrow Tory majority to govern for five years according to the programme for government they'd put forwards. Both the narrow majority and the programme in question were inconvenient for the Government, so they sought to overturn that result opportunistically.
Well sure, just as Labour now want to opportunistically overturn the 2017 result and have a snap election.
IIRC all the polls in a Mayor of London election underestimated Boris Johnsons vote share. The one at the top was the closest but would otherwise be seen as an outlier. Taking an average is a very risky strategy because all samples have biases and pollsters have to use their skill and knowledge to mitigate those biaises. Martin Boon of ICM famously came a cropper because he discredited the Labour vote share in the last GE as "the Labour vote is always exaggerated." That seemed to me to be herding rather than the empirical research he was paid to carry out. Basically there is no substitute for understanding your data.
I’ll graciously accept your verdict on the pre-eminence of Physics (though I did it as the easy option). I’d like to add to your point about GCSEs (or their Scottish equivalents) and say that 5A*s or so is the point where they start thinking about if there are any reasons for not accepting a student. I’ve seen students I thought were dead certs not get an offer though.
I will admit I still think the worst mistake I ever made in my life was not to do Physics A-level. I was good at it and didn't find it hard. It is highly prestigious and eminently valuable, as of course is a physics degree (which is why of course you are such a rara Avis these days).
I was asked for advice by the very able son of a colleague on his A-levels. He loves History and is doing it for A-level, but he also loved Physics. He asked me if they went together. I replied I thought that it could be an error to see which ones went 'together' and it was as well to keep a spread. But, I added, nobody will lose out in life from having a physics a-level. So he's doing that too.
The stuff about 7s I had from an Oxford admissions tutor. How truthful she was being I don't know, but it seems plausible, and with the effective end of AS levels (which actually I think was one of Gove's better ideas) it will only become more so.
I took Maths, Physics, Chemistry and English at A Level (along with the old style S levels in Maths and Physics). My version of your worst mistake was taking English Literature at university....
Unless students are certain what they want to do at university, I wouldn’t recommend mixing A Levels like that.
How - how could people be 'cheated out of Brexit' by another referendum? It'd need the people to vote to Remain after all, which they won't do if they want Brexit after all. Were the people 'cheated out of' a Conservative majority last year?
Because we haven’t implemented the result of the last one yet...
Like we'd implemented the entire 2015 Tory manifesto? No way. Obviously, the 2017 GE did not respect the result of the 2015 GE and calling it was democratically illegitimate.
Opportunistic, unwise and unnecessary certainly but how can holding an election be democratically illegitimate?
No more than holding a referendum is or is not so. I agree. I was employing sarcasm.
Holding a new referendum after we've exited the EU is entirely reasonable.
As was holding a new election after the last election's winners had formed a government.
Holding a referendum now would be like Corbyn winning the 2017 election and then protestors saying "lets have another election to check you really wanted Corbyn in power" before he entered Downing Street.
We voted to Leave, lets Leave and then we can talk about new votes.
Yeah, sure. Convenient definition you've invented. A General Election vote expires the moment a government is formed. Seriously?
A democracy that can't change its mind ceases to be a democracy. You're fine with the will of the majority when it marches with what you want, but now you suspect the majority is against that, nope, we don't want that, you can't change your mind until all of it is complete. (Thank God you're not a pilot: "Sorry, I don't care if it's looking really dicey, we've decided to land, I'm not aborting the landing, if we want to go around, we'll just have to wait until we've landed first and then take off again") And, preferably, with extra hurdles put in place to change one's mind (it's notable how many Leavers who say "voting to re-enter would be fine" and then add something like "of course, given that we'd have to accept extra this, that, and the other, I don't see it happening")
IIRC all the polls in a Mayor of London election underestimated Boris Johnsons vote share. The one at the top was the closest but would otherwise be seen as an outlier. Taking an average is a very risky strategy because all samples have biases and pollsters have to use their skill and knowledge to mitigate those biaises. Martin Boon of ICM famously came a cropper because he discredited the Labour vote share in the last GE as "the Labour vote is always exaggerated." That seemed to me to be herding rather than the empirical research he was paid to carry out.
Martin never herds.
Reasons why ICM were the top pollster in 1997, 2001, and 2010.
How - how could people be 'cheated out of Brexit' by another referendum? It'd need the people to vote to Remain after all, which they won't do if they want Brexit after all. Were the people 'cheated out of' a Conservative majority last year?
Because we haven’t implemented the result of the last one yet...
Like we'd implemented the entire 2015 Tory manifesto? No way. Obviously, the 2017 GE did not respect the result of the 2015 GE and calling it was democratically illegitimate.
The Tories had formed a government after the 2015 GE. The fact they decided to then go back to the polls earlier than expected is neither here nor there, the 2015 results had been honoured.
The 2016 results haven't been yet. We've not left yet. If we have another referendum after we leave but before all Vote Leave's promises have been honoured then that would be comparable.
The vote was for a narrow Tory majority to govern for five years according to the programme for government they'd put forwards. Both the narrow majority and the programme in question were inconvenient for the Government, so they sought to overturn that result opportunistically.
Well sure, just as Labour now want to opportunistically overturn the 2017 result and have a snap election.
Had the Entrance Exam disappeared by 1992? I suspect it had.
The Cambridge Sixth Term Entrance Paper (known to generations of students as STEP) is still used for Maths, usually on top of an A* or two at A-level.
Correct me if I'm wrong, but didn't they go (to be replaced just by interviews) and then come back?
I’m fairly certain the Maths ones never went away. I’ve taught a lot of students who had offers like A*A*AA STEP 1.
They were less than impressed when I told them what my offers were back in the day: two Cs from Imperial being the highest...
My Med School offer was BBC, though I got AAAB, similar to my peers. It gave a margin of safety when facing finals.
May I ask when that was? Are you referring to pre-late 1980s A levels? The assessment system changed radically thereafter - and grade inflation became rampant!
1983.
@Fysicsteacher is right about predicted grades. Private schools notoriously over egg them, some more than others. One of the most effective way my University has widened access fairly is by making post results offers. Basing on actual rather than predicted results has substantially increased the number of successful state school admissions. The extra year gives added maturity too.
I would think that the present day equivalent of BBC in 1983 would be circa AAA.
Comments
They made the same offer to everyone in the room: I’m not going to disagree.
Edit - on your substantive point, I can't answer for Maths, but I didn't have to sit an exam for History and Economics.
The idea that a poll has an error of +/- 3% has been invalid for many years now. It hasn't been used in US polls recently and the Sturgis report recommended[1] that it no longer be used. The BPC agreed and issued a new rule[2] on 2018-05-01 saying that it should be replaced by this statement.
"All polls are subject to a wide range of potential sources of error. On the basis of the historical record of the polls at recent general elections, there is a 9 in 10 chance that the true value of a party’s support lies within 4 points of the estimates provided by this poll, and a 2 in 3 chance that they lie within 2 points."
[1] http://eprints.ncrm.ac.uk/3789/1/Report_final_revised.pdf . See pages 76 and 77
[2] http://www.britishpollingcouncil.org/british-polling-council-introduces-new-rule-on-uncertainty-attached-to-polls/
Voters hold TM in far higher regard as a leader and a person than Boris.
Poll details are terrible for Boris supporters
If he hadn't had one good subject, would his results have been cruelly spelled 'Dunce?'
My brother got into LSE with A levels in Maths, Physics and Economics, but it was 1981!
So probably a number between 2% and 4%?
We are talking about best estimates.
If you don't know whether pollster X or Y has systematic bias, then the average of the two is the best estimate. Obviously if you do know which has no bias, that's the one to follow. But you don't know.
All pollsters are shooting at a target of being the most accurate. As you say they use different methodologies. But averaging them, in the absence of knowledge of which is the most accurate methodology, produces the best estimate. To say this is "Frankenstein nonsense" is to use emotive language to cover up a lack of understanding.
You say "There is plenty of evidence that the biggest source of error is systematic." Can you point to that evidence? Do you agree that the random error is about +/- 3% about 95% of the time? 5% of the time you get outliers outside that range. Do you think that the systematic error of pollsters is generally greater than 3%? If so, why?
As an aside, I have a school friend who went to Bath to do economics. I think she did OK for herself in the end. She's now a Professor at some small university in the USA, called Stamford Business School I think. Some of you may have heard of her as well - her name is Alison Goldsworthy and she's a former senior Liberal Democrat student activist, one of the women who had the misfortune to be involved in the Rennard case.
Yayes to confirmation 50
Nayes 48
The Senate has therefore voted to confirm Kavanaugh to be appointed as a US Supreme Court Justice
One of them says 11:55.
Another says 12:05.
What's the right time?
The average of two polls is not necessarily more accurate than one of the polls. In the EU referendum the online polls were more accurate than the the telephone polls and the average of them was not the closest to the truth. When the experts looked at this (eg [3]) they were all over the place.
The average is closer to the truth in the case where all the samples are done using the same method at the same time from the same sample frame, providing the method is random and the frame representative. In that case the Central Limit Theorem kicks in and the samples vary normally around the truth, so a poll-of-polls has merit.
But IRL the samples are nonrandom, done using different methods, the companies use different weights and different sample frames, and VI changes over time. The CRT doesn't apply here and there's no theoretical justification for assuming a poll-of-polls is better.
[3] https://www.populus.co.uk/wp-content/uploads/2016/03/Polls-Apart-29-March-2016.pdf
I do not know how they get away with it
US Senate voting is just crazy. I am not sure what the 30 hour delay was all about. It didn't change a thing. Allowing senators to change their votes is crazy.
It is no way to run anything.
STDs are increasing (although the group where they are rising fastest is not the young, but the old), but that's not because people are having more sexual partners, but because we've all stopped worrying about AIDS.
"The inquiry was also tasked with assessing whether uncertainty in poll estimates is
adequately communicated to stakeholders and the public. Our conclusion is that it is
not. Media commentators generally use a rule of thumb that point estimates of party
shares have a margin of error of + or - three percentage points. This is (or appears to
be) based on the calculation of a 95% confidence interval for a simple random sample of
1000 respondents for a party with 50% of the vote. But polls are not simple random
samples, parties are typically far from 50% of the vote, and when the full sample is 1000
the number of respondents used for estimating party shares is much less than this
number. For these reasons, the true margin of error will usually be different from + or –
3 per cent, sometimes larger and sometimes smaller, depending on the specifics of the
research design. "
Much of the terms used in polls ("margin of error", "standard deviation", "confidence interval") have their roots in statistical theory. But the compromises required to get a poll out quickly and cheaply (nonrandom sampling, nonrepresentative models, turnout adjustments) push the theory beyond breaking point. This is one on the reasons that recommendation was based on past results instead of a theoretical formula: the theory doesn't apply any more and hasn't done for many years now.
https://twitter.com/EuroGuido/status/1048629017719582725
https://twitter.com/EuroGuido/status/1048629868030808064
I know the average of two polls is not necessarily more accurate than one of the polls. We all know that. But we don't know which poll! That's the point.
I know all about the Central Limit Theorem and that the theoretical conditions don't apply with mixed polling methodologies. But I'm a practical mathematician not a pure one, trying to get a feel of what is the likely shape of the next government based on current polls, rather producing a water tight statistical paper than will pass peer review.
A new survey finds 55 percent of Americans regularly take a prescription medicine—and they're taking more than ever.
https://www.webmd.com/drug-medication/news/20170803/americans-taking-more-prescription-drugs-than-ever-survey
One in ten have illegal drug abuse problem.
https://www.northpointwashington.com/blog/percentage-population-struggles-addiction-answer-may-surprise/
That’s not the figure of everyone using illegal drugs, if for now they don’t have a problem that’s not to say it won’t ever be because the nature of addiction is need for more just to stand still. Also of course if something is against the law, and a sackable offence for employers, how do you accurately measure what is really going on?
As you state, the trend may not seem bad to you, but to me these are very high and very sad %s. If it’s about impacts, then surely such things can break relationships, families, wreck homes, break people? And that means in the bigger picture, problems for society with crime, state support, mental health.
For GB I suspect lower %s than US, but If it’s about trends, then surely trend in US and GB is in the wrong direction? Here in GB, police are saying to middle class drug users, you are fuelling crime epidemic, are they not?
Here’s an STD stat. 25 percent of Americans have an incurable sexually transmitted disease.
https://edition.cnn.com/2018/08/28/health/std-rates-united-states-2018-bn/index.html
I say if true that’s shockingly high. Anybody want to disagree?
Some dismiss my posts as unserious others claim I shouldn’t be called LordOfReason. But I think to ask is it liberal politicians or Conservative politicians who have the values to bring these %s down is a reasonable question to ask.
The biggest fundamental problem to me is probably differential non response, closely followed by people outright lieing to the pollster with people changing their god damned minds at the ballot box being another issue.
@Fysicsteacher is right about predicted grades. Private schools notoriously over egg them, some more than others. One of the most effective way my University has widened access fairly is by making post results offers. Basing on actual rather than predicted results has substantially increased the number of successful state school admissions. The extra year gives added maturity too.
Yougov's 'map' seemed to work well for 2017, hopefully they'll do it again.
You can just use your gut of course.
Guido's as bad at cherry-picking and repackaging facts to make a point as anyone else so it's an act of hypocrisy on his part when he calls people out on it. It's just frustrating that the entire commentariat seem to be this bad and twitter mayhaps has led to further degeneration.
If professional political commentary roles were allocated on the grounds of quality and insight, David Herdson would be one of this country's foremost columnists and talking heads.
Wow.
OK, even that statement is wrong (as the EU ref proved) but it's a bit more reliable.
No way. Obviously, the 2017 GE did not respect the result of the 2015 GE and calling it was democratically illegitimate.
The 2016 results haven't been yet. We've not left yet. If we have another referendum after we leave but before all Vote Leave's promises have been honoured then that would be comparable.
And implemented a popular part of their manifesto - an in/out referendum.
I agree. I was employing sarcasm.
Do you agree these are massively high stats for drug problem and STD rates?
is it liberal politicians and judges or Conservative politicians and judges who have the values to bring these %s down would you say?
As was holding a new election after the last election's winners had formed a government.
Holding a referendum now would be like Corbyn winning the 2017 election and then protestors saying "lets have another election to check you really wanted Corbyn in power" before he entered Downing Street.
We voted to Leave, lets Leave and then we can talk about new votes.
Once we've left we can have another vote. Even if its opportunistic. We haven't left yet so its not comparable.
........
https://twitter.com/hendopolis/status/1048676793283223552
Unless students are certain what they want to do at university, I wouldn’t recommend mixing A Levels like that.
A democracy that can't change its mind ceases to be a democracy.
You're fine with the will of the majority when it marches with what you want, but now you suspect the majority is against that, nope, we don't want that, you can't change your mind until all of it is complete. (Thank God you're not a pilot: "Sorry, I don't care if it's looking really dicey, we've decided to land, I'm not aborting the landing, if we want to go around, we'll just have to wait until we've landed first and then take off again")
And, preferably, with extra hurdles put in place to change one's mind (it's notable how many Leavers who say "voting to re-enter would be fine" and then add something like "of course, given that we'd have to accept extra this, that, and the other, I don't see it happening")
Reasons why ICM were the top pollster in 1997, 2001, and 2010.
If a GE is fine, so is a referendum.
If we get lucky we can give Northern Ireland away to the Republic or the EU as part of the Withdrawal Agreement.
https://twitter.com/paulhutcheon/status/1048679816881471495