Please register to participate in our discussions with 2 million other members - it's free and quick! Some forums can only be seen by registered members. After you create your account, you'll be able to customize options and access all our 15,000 new posts/day with fewer ads.
Most polls these days randomly ask less than 1% of the public in order to gauge the current opinion of the voters, whether or not the people asked are even voters. Wouldn’t it be more accurate if they could maybe ask 50% of actual voters to see what they think?
Didn’t that happen in November 2016? Even if 50% of eligible voters didn’t bother to vote the result is far more accurate in determining the will of all the people than a random telephone call could ever determine.
In 2016, enough people were unhappy with the direction the previous politicians were heading in, such that a president was elected who espoused policies in the opposite direction. Since then the same people who lost in that election have pushed even farther in the same direction, while a significant number of people who had supported their views at the time have been realizing improvements in their situation under the new president.
I wonder if the descendants of Thomas Paine have a copyright on the title “Common Sense”. We could sure use a new version of that work.
Most polls these days randomly ask less than 1% of the public in order to gauge the current opinion of the voters, whether or not the people asked are even voters. Wouldn’t it be more accurate if they could maybe ask 50% of actual voters to see what they think?...
Try 0.00031%. For the national pollsters. A number so statistically small that no scientific method would rely on it. i.e. Polling is a pseudo science.
Try 0.00031%. For the national pollsters. A number so statistically small that no scientific method would rely on it. i.e. Polling is a pseudo science.
All the mathematical principles of geometry are fixed. The math has not changed for centuries, because it has always proven to be accurate. So it is with statistics. If the math is accurate, the results are always the same.
When statistics are applied correctly, a sample size can be quite accurate. The size doesn't depend on large numbers for accuracy in medicine, economics, management, military, and all the other areas where statistics are always used.
Sample size in statistics is like the sides of a triangle in geometry. The triangle doesn't have to be huge to determine the correct angles. The sample size doesn't need to be huge to be accurate in statistics. Polling is nothing but applied statistics, just like building a house is applied geometry.
But polling is people, and geometry is objects. People's minds change much more often than objects change. So polling is more fluid, and depends more on the fluid medical science of psychology than only statistics. Psychology finds the right questions to ask. Statistics verifies the validity of the questions.
The sample has to be carefully constructed to begin with. Once the hypothesis is carefully examined, the size of the sample's accuracy can then be determined. A large sample will only add repetition to the results, not increased accuracy.
However, if there is something wrong is suspected with the hypothesis, sometimes a larger sample is taken to ensure a mistake wasn't made. But mistakes are always accounted for in the original hypothesis, and mistakes are always a part of statistics. They are accounted for right from the beginning of the process.
Since polling is all about statistics, the hypothesis is always the first thing that needs to be examined. Polling can always have an agenda behind it, and if an agenda exists, the poll can always find the results the polling questions if the hypothesis is skewed.
That's why legitimate pollsters always do predictive polling with a given range of inaccuracy. Typically, this range is +5% to -5%, but that varies, often downward, as the pollsters collect more data.
But questioning 100 people can be just as accurate as questioning 1000 people or 10000 people if the questions are well crafted and the people are demographically accurate. The first questions of any good poll are used to determine how accurate the following questions will be.
So is polling a pseudo-science? No. But it can be used by hucksters in an illegitimate way, which is psuedo-scientific; it can appear to be scientific to the ignorant observer, but to one who knows the science, it's phony.
And since statistics is something most folks don't study very deeply, it's very easy for a company with an agenda to push to create phony polls. That's why studying an aggregate of polls is the only easy way the average person can determine what polls are accurate or not.
The outliers are almost always those with agendas they want to see accomplished for their benefit. Legitimate pollsters are quite willing to report poll results that are unfavorable to someone or something.
The object of political polling is to determine how good or poorly a candidate is doing at the moment. Candidates pay pollsters for this information.
The wise candidate takes an unfavorable poll to heart and tries to change things in the campaign to make the next poll more favorable.
The stupid candidate rejects the poll's findings and continues on unchanged.
The candidate's followers are the same. Some will see their candidate is in trouble and some will stick by their guy, right or wrong.
But when an aggregate of all the polls is taken and averaged, if a candidate is in trouble, he's in trouble for a fact.
Belief does not change facts. For every pollster who asks the wrong questions, there is another who asks all the correct questions.
When 3 different doctors all give their diagnosis, and the patient hears all 3, that's polling. The patient can always find a 4th, 5th, 6th, or 7th doctor for more diagnoses.
When all the doctor's diagnoses except for 1 doctor indicate cancer, statistics say the patient has cancer. Believing all the other doctors are wrong doesn't stop the cancer, even if the patient believes in the single doctor who disagreed with the others.
Last edited by banjomike; 07-11-2019 at 10:05 AM..
All the mathematical principles of geometry are fixed. The math has not changed for centuries, because it has always proven to be accurate. So it is with statistics. If the math is accurate, the results are always the same.
When statistics are applied correctly, a sample size can be quite accurate. The size doesn't depend on large numbers for accuracy in medicine, economics, management, military, and all the other areas where statistics are always used.
Sample size in statistics is like the sides of a triangle in geometry. The triangle doesn't have to be huge to determine the correct angles. The sample size doesn't need to be huge to be accurate in statistics. Polling is nothing but applied statistics, just like building a house is applied geometry.
The sample has to be carefully constructed to begin with. Once the hypothesis is carefully examined, the size of the sample's accuracy can then be determined. A large sample will only add repetition to the results, not increased accuracy.
However, if there is something wrong is suspected with the hypothesis, sometimes a larger sample is taken to ensure a mistake wasn't made. But mistakes are always accounted for in the original hypothesis, and mistakes are always a part of statistics. They are accounted for right from the beginning of the process.
Since polling is all about statistics, the hypothesis is always the first thing that needs to be examined. Polling can always have an agenda behind it, and if an agenda exists, the poll can always find the results the polling questions if the hypothesis is skewed.
That's why legitimate pollsters always do predictive polling with a given range of inaccuracy. Typically, this range is +5% to -5%, but that varies, often downward, as the pollsters collect more data.
But questioning 100 people can be just as accurate as questioning 1000 people or 10000 people if the questions are well crafted and the people are demographically accurate. The first questions of any good poll are used to determine how accurate the following questions will be.
You are waaaaay too logical for this place. Please realize that most don't have the foggiest notion of statistics.
The reality is that even for a very large population, there is no significant increase in accuracy gained by increasing the sample size - given the assumptions you made above.
Here is a pretty good article on the topic by Scientific American.
One of the most important observations they make in the article is that "a poll is a snapshot, not a forecast". This is because while the poll may have an accuracy +/- 3% at any given moment in time, opinions can change rapidly. Which is why of course political polls are repeated frequently.
Most polls these days randomly ask less than 1% of the public in order to gauge the current opinion of the voters, whether or not the people asked are even voters. Wouldn’t it be more accurate if they could maybe ask 50% of actual voters to see what they think?
The math is solid, it's not a matter of opinion. Rigorous math shows that no matter what you are sampling -- whether it's voters, beans, or molecules -- that if you have 100 million of something you only need to sample about 1,000 of them to have a margin of error of 3% with 95% confidence.
From this table you can see why polls usually ask about 1,000 respondents. Polls would be more accurate if they polled 15,000 people, but that would take considerable more amount of time and money, so they don't. Good polling is an art because the sample has to be balanced to represent all areas of the country, ages, races, parties, etc.
All the mathematical principles of geometry are fixed. The math has not changed for centuries, because it has always proven to be accurate. So it is with statistics. If the math is accurate, the results are always the same.
When statistics are applied correctly, a sample size can be quite accurate. The size doesn't depend on large numbers for accuracy in medicine, economics, management, military, and all the other areas where statistics are always used.
Sample size in statistics is like the sides of a triangle in geometry. The triangle doesn't have to be huge to determine the correct angles. The sample size doesn't need to be huge to be accurate in statistics. Polling is nothing but applied statistics, just like building a house is applied geometry.
But polling is people, and geometry is objects. People's minds change much more often than objects change. So polling is more fluid, and depends more on the fluid medical science of psychology than only statistics. Psychology finds the right questions to ask. Statistics verifies the validity of the questions.
The sample has to be carefully constructed to begin with. Once the hypothesis is carefully examined, the size of the sample's accuracy can then be determined. A large sample will only add repetition to the results, not increased accuracy.
However, if there is something wrong is suspected with the hypothesis, sometimes a larger sample is taken to ensure a mistake wasn't made. But mistakes are always accounted for in the original hypothesis, and mistakes are always a part of statistics. They are accounted for right from the beginning of the process.
Since polling is all about statistics, the hypothesis is always the first thing that needs to be examined. Polling can always have an agenda behind it, and if an agenda exists, the poll can always find the results the polling questions if the hypothesis is skewed.
That's why legitimate pollsters always do predictive polling with a given range of inaccuracy. Typically, this range is +5% to -5%, but that varies, often downward, as the pollsters collect more data.
But questioning 100 people can be just as accurate as questioning 1000 people or 10000 people if the questions are well crafted and the people are demographically accurate. The first questions of any good poll are used to determine how accurate the following questions will be.
So is polling a pseudo-science? No. But it can be used by hucksters in an illegitimate way, which is psuedo-scientific; it can appear to be scientific to the ignorant observer, but to one who knows the science, it's phony.
And since statistics is something most folks don't study very deeply, it's very easy for a company with an agenda to push to create phony polls. That's why studying an aggregate of polls is the only easy way the average person can determine what polls are accurate or not.
The outliers are almost always those with agendas they want to see accomplished for their benefit. Legitimate pollsters are quite willing to report poll results that are unfavorable to someone or something.
The object of political polling is to determine how good or poorly a candidate is doing at the moment. Candidates pay pollsters for this information.
The wise candidate takes an unfavorable poll to heart and tries to change things in the campaign to make the next poll more favorable.
The stupid candidate rejects the poll's findings and continues on unchanged.
The candidate's followers are the same. Some will see their candidate is in trouble and some will stick by their guy, right or wrong.
But when an aggregate of all the polls is taken and averaged, if a candidate is in trouble, he's in trouble for a fact.
Belief does not change facts. For every pollster who asks the wrong questions, there is another who asks all the correct questions.
When 3 different doctors all give their diagnosis, and the patient hears all 3, that's polling. The patient can always find a 4th, 5th, 6th, or 7th doctor for more diagnoses.
When all the doctor's diagnoses except for 1 doctor indicate cancer, statistics say the patient has cancer. Believing all the other doctors are wrong doesn't stop the cancer, even if the patient believes in the single doctor who disagreed with the others.
I can’t rep you, but that was a great post, even if it will be totally ignored by the people who don’t want to understand how polling works, and would rather claim that all polls are biased unless they have the exact number of Republicans and Democrats in the sample.
Aside from the math lesson, how about the part about the losers pushing an agenda that is even further to the extreme than they lost on in 2016? Can anyone agree that the election represents a pretty good poll by itself?
Last edited by Cruzincat; 07-11-2019 at 10:32 AM..
You are waaaaay too logical for this place. Please realize that most don't have the foggiest notion of statistics.
The reality is that even for a very large population, there is no significant increase in accuracy gained by increasing the sample size - given the assumptions you made above.
Here is a pretty good article on the topic by Scientific American.
One of the most important observations they make in the article is that "a poll is a snapshot, not a forecast". This is because while the poll may have an accuracy +/- 3% at any given moment in time, opinions can change rapidly. Which is why of course political polls are repeated frequently.
Yup.
As soon as any snapshot is taken, someone it it moves.
Frequency is paramount in political polling. Last week is ancient history in any political campaign as an election nears the end.
Polling can be maddening to keep up with, but that never means the pollsters are wrong.
People just don't like discouragement when their guy is showing he's in trouble.
Please register to post and access all features of our very popular forum. It is free and quick. Over $68,000 in prizes has already been given out to active posters on our forum. Additional giveaways are planned.
Detailed information about all U.S. cities, counties, and zip codes on our site: City-data.com.
Please register to participate in our discussions with 2 million other members - it's free and quick! Some forums can only be seen by registered members. After you create your account, you'll be able to customize options and access all our 15,000 new posts/day with fewer ads.
