Any statistics experts?

Discussion in 'Politics, Religion, Social Issues' started by Neserk, Jul 8, 2004.

  1. Neserk macrumors 6502a

    Neserk

    Joined:
    Jan 1, 2004
    #1
    Okay, I took statistics in college. I actually understood it and did well. I can't remember something though...

    How many people do you need to get a good representative sample of the US population?

    I'm seeing polls all over the place and when I read that they only polled 1,000 people it makes me suspiciuos. Is that enough?

    I also wonder where they randomly took their sample from... but that is a different matter!
     
  2. Thanatoast macrumors 6502a

    Thanatoast

    Joined:
    Dec 3, 2002
    Location:
    Denver
    #2
    It is around 1000 people, but 1500 is better. It also depends on how the poll was conducted, and who was polled. Polling voters, as opposed to likely voters, or even randomly will all generate different results. Rolling polls, taken over serveral days will also generate better results. Another good way to look at the validity of the poll is to check the margin of error. Anything over 3.5%, should be taken with a grain of salt. Also, reputable polling houses are the best. (But I've forgotten which ones are generally well-respected) - Gallup, I think is fairly safe. Also, if they print it, read the question they asked.
     
  3. mactastic macrumors 68040

    mactastic

    Joined:
    Apr 24, 2003
    Location:
    Colly-fornia
    #3
    IIRC 1000 will give you a +/- of 3.5%. More people and the margin of error drops, but it takes a whole lot of people to get it to drop significantly (ie. below 2% +/-) therefore from a cost/benefit standpoint 1000 people is a good balance between accuracy and cost.

    Plus with all the telemarketing barrage, not to mention people screening calls, avoiding calls with CID, etc, it gets harder and harder to reach those 1000 people all the time. I forget the numbers, but I was just reading something saying that you have to call something like 3 times as many people as you did 10 years ago to get the same number of respondents.

    Anyway, my opinion of polls is still the same. When asked, lie like a rug. If enough people do this and polls are proven wrong on enough occasions, perhaps polititians will have to go back to the days of basing an opinion on personal conviction rather than the direction of the wind... although I hear that's supposedly only a liberal trait.
     
  4. Voltron macrumors newbie

    Voltron

    Joined:
    May 9, 2004
    #4
    Thus the only opinions they get are those too stupid or too poor to screen their calls?
     
  5. mactastic macrumors 68040

    mactastic

    Joined:
    Apr 24, 2003
    Location:
    Colly-fornia
    #5
    Well no, another whole area of discussion is their representative sample but the details of that were unimportant to what Neserk was asking. I'm also not saying that only stupid people answer, indeed many people view answering polling a civic duty. Not everyone wants to screen those calls. I'm just saying with the rise in that technology, there is a corresponding decrease in number of people who respond to the cold calls.

    'They' have to call X number of people from demographic A and Y number of people from demographic B etc. You can't just go through the phone book and dial for these kinds of polls. It is very complicated, and many advanced degrees have gone into what the best way of assembling a sample that represents the population adequately enough to make assumptions based on it.
     

Share This Page