Edit: had no idea “poll” was such a four letter word, especially when talking about them in the abstract. Anyone want to chime in on the downvotes? Is it just “all polls are bad” or is it Nate Silver? Honestly had no idea talking about poll weighting would be so unpopular.


Since there’s always a lot of interest in the validity of polls, I found this to be interesting. It’s Nate Silver’s explanation of how they do weighting of polls when aggregating based on the pollsters track record. He makes it clear that the bias is often a result of the methodology and not necessarily a “thumb on the scale” and how the pollster executes a poll can introduce bias - and how they account for that.

Many folks have issues with Nate, but he’s at least very transparent in how they account for bias based on previous performance, not just the poll source. So while you may disagree with his decisions, you can at least look at his numbers and know how they got there.

  • geekwithsoul@lemm.eeOP
    link
    fedilink
    English
    arrow-up
    3
    ·
    2 months ago

    I am just as confused. Most of my /politics posts and comments are relatively well upvoted, but I guess there’s a sizable contingent who just hate any mention of polls at all or have it out for Nate Silver. I was at least expecting a few more comments about how Silver ranked some pollsters and whether folks thought those assessments were accurate. 🤷🏼‍♂️

      • geekwithsoul@lemm.eeOP
        link
        fedilink
        English
        arrow-up
        3
        ·
        2 months ago

        In general I don’t think so, and there’s only about one to three accounts that downvote most things I post/comment so I don’t think it’s that. I think I’m leaning towards reactionary downvotes for anything that mentions “polls” - which is a shame, as while I don’t think polls are terribly good at being reliably predictive of the results, I think they’re fairly good when used for tracking changes in momentum and at least more reliable when analyzed in aggregate with competent and rigorous weighting based on past performance. And this article is great about explaining how that’s done by the only folks who’ve been shown to be any good at it.