Please Explain This To Me.
Submitted by andyram on Thu, 01/03/2008 - 10:21
Okay.. all of these polls have a e.g. 4.5% or some sort of margin for error... If they poll 800 people, where do they get this margin of error? Do they just guess since everybody isnt polled? I dont understand this.. Where do they get this margin of error number? I know polls are about as useful as a Federal Reserve Note.. but I am just curious..
»
Another thing...
One of the things that these polls cannot factor in is those voters who only have cell phones.
They only call those with the standard "copper line" house phone - and that is only to those who have been involved in primaries or votes in the past.
The polling system became defunct in the age of digital/wireless communications. They have no way to get accurate data, no matter what they say about their statistics.
We are in a digital "Battle of Thermopylae."
Persians (Old Media): Our media coverage makes the world move according to our wishes. Our corporate owners control every facet of American life. Our profiteering from wars fill our accounts with unlimited funds to crush your petty little "r3volution." Our slanderous attacks against you on the airwaves will blot out the sun...
Spartan King (Dr.Paul): Then we shall fight in the shade...
Until the Election is Won!
Chaplain Steve
Ron Paul - Lion of the Constitution
Until the Election is Won!
Chaplain Steve
Ron Paul - Lion of the Constitution
Life lessons from The Simpsons
"Oh, people can come up with statistics to prove anything, Kent. 40% percent of people know that."
- Homer
margin of error
The margin of error is based upon the sample size vs. the total target voting population it is projecting. It assumes the sample polled is a completely random sample of the target voting population being polled. It also assumes the polling sample accurately includes all the voters that will be in the total voting population, while excluding people that will not be voting. If these are correct then the margin of error is the range that mathematically the poll result will reflect the final voting result from the total population 95% of the time.
To be accurate, however, the poll sample must be completely random, and it must be taken from the entire population that will be voting, and it must not include samples of people that will not be voting.
As has been pointed out in many places, however, the polls do not do this. Because of the methods used to conduct the polls (i.e. landlines and only people who answer and respond to such calls) they are not random. Also, because the pollsters focus so hard trying to exclude people that are not likely to vote, they are likely in this election to be excluding many likely voters from their polls. For instance they may exclude people who have never voted before or have not voted recently. They may exclude people who are not Republican or are only recent Republicans. They may exclude people who have not voted in the most recent Republican primaries or the last Republican presidential primary. They may also weigh the results of each person polled (i.e. counting as 1.5 votes in their calculation v.s 1 vote) in an attempt to assess the actual likelihood of voting (i.e. long time voter vs. new voter).
All of this means that they are not sampling the entire population of the voters, and that the samples they do poll are not random but skewed toward those with landlines who actually answer these poll solicitation calls and take the time to repond to the questions, as well as skewed towards other factors.
There is also pressure on professional pollsters to have results very similar to other polls. If a pollster has a result that greatly differs from the rest of the polls and that pollster turns out to be wrong. That pollster has just lost all credibility and may go out of business. However if the pollster has similar results as all the other polls and they all turn out to be wrong, well then it as something completely out of their control - factors that could not be accounted for.
Let it not be said that we did nothing.-Ron Paul
Stand up for what you believe in, even if you stand alone.-Sophia Magdalena Scholl
POLL
maryj
THEY NEVER GET THE TRUE RESULTS FROM A POLL AND THEY DO NOT KNOW WHY.
SO THEY COVER UP NOT KNOWING HOW TO POLL AND SAY THEY COULD HAVE MADE AN ERROR.
Liberty is our campaigns gift to the world it is our manifest destiny.
Freedom is a movement who's time has come. Our campaign will lead world wide freedom.
Statistics 101
The act of polling 800 people is supposed to be a representative sampling of a larger population. When you get the results among the sample population (800 people), you use that to try to predict or extrapolate the results among the larger population. The margin for error does not apply to the sample population. You can never be 100% accurate because it is simply not feasible (normally) to sample the entire population (could be city, state, region, nation), therefore you use statistical formulas to figure a "margin for error" among the larger population. I.e., if 800 people were polled and 80 of them voted for Ron Paul (10%), the prediction would be that the larger population of 80,000 (for argument's sake) would also vote for Ron Paul, with a "margin for error of about 4.5%". This means that as many as 11,600 or as little as 4400 people would actually vote for Ron Paul. My example has quite a wide margin for error I know, real polls are designed to narrow the spread as much as possible, statistically speaking of course.
However, as we have seen all during 2007, when you include "bias" in your sampling methods (such as excluding candidates names and removing people's phone numbers when they select "other"), you can pretty much throw those polls out as pure bunk. Unfortunately that is the bane of statistics, they can be used to paint any picture you want.
I got straight A's in all my business stats classes I took many years ago, hopefully it all came out right. I think the basics are there.
=======
RON PAUL 2012
Thank You
...for the knowledgeable insight.
Margin of error link
http://en.wikipedia.org/wiki/Margin_of_error
Here is a wikipedia link that explains margin of error in random samples. The problem with the current political polls are the actual samples they are working from. To be an accurate survey they must be as random as possible with no bias. This is why internet polls are meaningless. They are not random and only highly biased peopel are motivated to vote in them. The current random polling is suspect however as we know sampling "likely Republican primary voters" culled from the heap of last elections voters will simply not produce good numbers for Ron Paul as his supporters tend to be independents, first time voters and even people who voted Democrat or third party last go around. Simply put Ron Paul is a new political animal that the old school pollsters are not taking into account. When you combine this with the old school methodology such as not calling cell phone users and even not including candidates in the questionaire well you get inaccurate results. This is not to say that Ron Paul is getting HUGE amounts of support more than the conventional polls are indicating but I believe it is a safe assumption that he is getting significantly more support than those polls are indicating.
This number comes from
This number comes from standard statistics. It's not actually a 4.5% margin of error. It's a 95% certainty that the error is within 4.5%. That kind of means that there is a 95% chance that if the same poll were done again that the difference would be no more than 4.5%. The percent error is very scientific and I don't like it when people always assume that we are at the upper end of the percent error. This may be true in one poll but couldn't be the case for very many polls so we can't use this argument every time. However all they are saying is that they sampled subgroup 'n' of group 'N'. Little n is the group of people that they sampled and big N is the total group of "likely Republican voters". In theory group 'N' is always the same size but each survey has a different subgroup 'n'. The bigger little n is the smaller the 95% certainty margin of error is. If they make the conditions of the survey such that group 'N' excludes many Ron Paul supporters (which I believe that they do either intentionally or unintentionally) then they can survey all day long and get their 95% errors down to 0.5% but that doesn't mean anything. Sorry for all of math stuff. My explaination is not perfect and I'm sure a statistician would shoot holes in it but I hope it gets the point across.
Thanks, that helps a lot!
Thanks, that helps a lot!
Lies, damn lies, and statistics!
I started to write a similar explanation, but I so deplore polls. I took two semesters of statistics at college and all I can really claim to have learned was that statistics can be manipulated to demonstrate anything. It is impossible to find a truly un-biased pollster these days. And the understanding of how to draw a truly representative sample pool is simply not to be found in any survey designs I have seen lately. Let you own eyes and ears be proof enough for you. Do you own informal poll whenever you have a moment when out in public. I'd trust your eyes and ears more than a published poll.
The meaning behind my screen name
The battle of Gonzales, TX. (Oct. 1835)
The problem is not with statistics...
It is with unethical (or ignorant) people CONSTRUCTING the statistics.
If you have a Box full of the same "widgets" and you grab a handful and weigh that handful then divide by the count of the handful to get an average weight, and then weigh the whole box & divide by the avg weight to get the count of widgets in the box, that is quite valid.
But people -- and especially voters -- are not widgets. You may THINK you know which people are "in the box" (i.e. people who will vote) but you really don't... because some people will decide at the last minute to jump into the box, and others will decide to NOT be in the box after all.
The actual "vote" on a given primary or caucus day is a "snapshot" of the contents of that box of SELF-SELECTED widgets. Sure there is SOME correlation of this election with past elections, but not much.
Why? Well, in a typical FINAL election, anywhere from 40 to 60% of a local population will "vote" -- but in a PRIMARY election, typically only 20% will participate (10% in each major party). Exactly WHO chooses to vote in any given primary, versus who decides the weather is too bad, or who "intends" to vote, but then forgets about it on their way home while picking up milk, diapers, etc --- AND how those who DO vote actually decide WHO to vote for, well that's really hard to define.
Polling is INTENTIONALLY biased -- by definition, it is attempting to use the "analyzed" biases of voters from past elections to predict the outcome of a future election. It is seen as valid because it often HAS worked (like maybe 80% of the time if you count "local" campaign races) -- but in national elections, the track record is ridiculously poor. (Remember, Howard Dean was the "inevitable" Democratic Nominee in 2004... until he wasn't, etc., ad nauseum)
Ignore the polls, and TELL PEOPLE to ignore the polls and just inform them about our candidate -- THIS YEAR there will likely be NO true "frontrunner" and people will finally have to make a REAL choice. Tell them NOT to do the "eeny-meeny-miney-moe" with their ballot (if they are gonna do that, better that they stay at home.)
Statistics
There actually is a science behind how to sample accurately, there are a number of mathematical tests based on the number of people you sample, and the size of the population, some examples - F-tests, T-tests, and others. By the way, they are only 95% certain that there number is within 3 -4 %.
That of course, does not consider the quality of their sampling mmethods - Land lines vs Door to door, mailing lists used, etc. And Of course, you have to assume that the people doing the sampling actually want an accurate number. If that is true, they probably can get within 10% or better of an accurate number. Then its about Turn out, Drew Ivers Claims that in a caucus an organization can make a 30 % difference.
don't forget
That the phrasing of questions has been proven (and perfected by dickheads like Luntz) to affect the outcome of a response. Answer responses have an order and a format, plus the order of the questions can affect the outcome as well. If you ask a string of questions about Iran and the nuclear threat, then ask which Rep candidate is best for protecting us, then the answer will be skewed to a McCain or Rudy since the respondent is in a mindset of 'Iran will kill me'
even the sex and racial tone of the interviewer can affect an outcome. Plus a whole host of intangibles such as time of day, frequency of respondent contact, etc. If you have been called 25 times in a day (like some ppl in Iowa are really experiencing today) would you sit and listen to any question or just say...'I'm voting for Huck, stop calling'?
Of course, good pollers avoid some of these pitfalls by randomizing question order and what not, but it is, like you mentioned, under the control of the poll makers what kind of response they want to get.
Polling can be accurate only under the right circumstances with the right questions. The only polls that mean anything are the exit polls and the final vote count.
-----------------------------------
I can know of no greater functional movement than a 600lb deadlift. Except a 700lb deadlift. Cuz when you are that strong, it's functional.
www.crossfit.com
-----------------------------------
I can know of no greater functional movement than a 600lb deadlift. Except a 700lb deadlift. Cuz when you are that strong, it's functional.
www.crossfit.com
Wik it!
http://en.wikipedia.org/wiki/Margin_of_error
Can't answer that but little
Can't answer that but little story about our poll experience. I am the registered republican in our house. My husband was home when we got the poll call ( it was automated anyway). Listed 5 candidates---Ron Paul NOT being among those & gave the option of pressing # 6 for " other ". Obviously, my husband pressed # 6 & the got the following message.
" Thank You, Your number has been permanently removed from this list. "
WTF!!! Is this supposed to be a good cross section of individuals?? Good thing medical studies aren't done that way.
Same thing
happened to me in Fla just after Thanksgiving. How many others??????
LOL
"I know polls are about as useful as a Federal Reserve Note.. but I am just curious."