• NoneOfUrBusiness@fedia.io
    link
    fedilink
    arrow-up
    5
    ·
    2 hours ago

    The poll they cite only polled 500 people

    Which is more than enough. Statistically, you only need 30 or so samples to get a normal distribution, at which case the error varies with 1/sqrt(n) for sample size n. This means that if a sample size of 8000 would only have 4 times less error (so from 4% to 2%) compared to one of 500. The mathematics of the thing is ironclad: Provided the sample is representative (i.e. doesn’t unfairly over or underrepresent a particular group) and exceeds the mid-double digits*, any issues with sample size would manifest as a large calculated error.

    • Jollyllama@lemmy.world
      link
      fedilink
      English
      arrow-up
      2
      ·
      1 hour ago

      Founded and funded by people and orgs that have a vested interest in this conflict. Like the heritage foundation polling for gay marriage.