Posts by Bart Janssen

  • Hard News: Where are all the polls at?, in reply to Dennis Frank,

    myth of scientific objectivity

    There have been studies that show pretty compellingly that people who believe themselves to be objective (eg scientists) are more subject to confirmation bias than average.

    However, when making comments on the meaning of data there is a difference between comments made from knowledge and comments made from ignorance. Both can be wrong but one is more likely to be wrong than the other. I rarely trust my accountant* to advise me on medical matters.

    *lol as if I'd have an accountant.

    Auckland • Since Nov 2006 • 4335 posts Report Reply

  • Speaker: Polling 2017: life beyond landlines?, in reply to Gavin White,

    if a poll shows NZ First (as the classic example) 1% below their final result that doesn't necessarily mean it was wrong

    Absolutely. 1% here or there is perfectly fine with this kind of sampling.

    The problem is we are seeing 10% differences and that realistically means one or more polls are very wrong.

    Auckland • Since Nov 2006 • 4335 posts Report Reply

  • Speaker: Polling 2017: life beyond landlines?,

    The key problem is these are samples that try to replicate the result you would get from polling every person who can vote (the actual election). The question is -
    "Is the sample accurate?"
    Accuracy is about how close the sample is to the "real" result of polling everyone.

    The problem we have is that we have one "real" result every election, and multiple samples (polls) each giving different answers. There are two options, either the samples are "real" and the population as a whole is changing it's opinion by huge jumps from poll to poll OR the samples are not accurate.

    Margins of error allow you to give a fudge around the number but the margins themselves can be fraught. But ultimately if your samples don't reflect the "real" result then you're doing it wrong.

    At present you have to conclude one or more of the polling companies is doing it wrong - pretty badly wrong as well.

    It makes for easy entertainment and fills minutes on the news - but personally I don't think it adds any value to our democracy and could well be doing harm.

    Auckland • Since Nov 2006 • 4335 posts Report Reply

  • Hard News: Where are all the polls at?,

    Just a comment about margins of error. At risk of 'splaining, most folks assume that the margin of error means the "real" result will be the stated number plus or minus the margin of error.

    But these errors have a bunch of assumptions and what they really mean is the "real" number is PROBABLY within that range but it could be well outside that given number.

    But there are other issues, a passing familiarity with stats should tell you that the margin should be smaller on the zero side. That is, if the number given is 20% then it should be +say 5% and minus say 3%, so right up front you know the margin of error is calculated in an unusual manner and you should treat it with suspicion.

    But the biggest problem with those margins is they make assumptions about the sampling. They assume the sample is random and the survey companies are struggling with that now. They also assume the samples have certain distributions (remember normal distributions from school?).

    Those assumptions are necessary to do simple stats but it's never certain the assumptions are valid.

    It's possible to do better statistics and my guess is internal party polling probably uses much more sophisticated models and error reporting but statistics makes most folks head hurt and the media is just plain reluctant to try and present anything complex.

    The result is, these entertainment polls with silly margins of error. They make for entertainment and sound bites and tweets but as for actually representing the opinion of the population - they're horrible.

    Auckland • Since Nov 2006 • 4335 posts Report Reply

  • Up Front: The Surprisingly Sincere Up…,

    I personally would probably drink less gin

    Why would you want to do that?

    Auckland • Since Nov 2006 • 4335 posts Report Reply

  • Hard News: Drugs and human rights, in reply to Tom Semmens,

    the dark hatred of objectifying human beings

    I know this is meant as an insult but it really is worth noting that in a neoliberal world, humans are just resources to be manipulated and spent to maximise profit.

    They really do believe humans are objects

    Auckland • Since Nov 2006 • 4335 posts Report Reply

  • Hard News: Drugs and human rights, in reply to Russell Brown,

    English seeks to walk it all back on Morning Report.

    As noted in the morning TV panel later on - "Even if she misspoke, if she said it, it's because she thought it"

    Auckland • Since Nov 2006 • 4335 posts Report Reply

  • Hard News: Where are all the polls at?, in reply to simon g,

    gave the Greens an extra point ... because I can!

    Since The Greens have polled from 15 to 4% and back to 6% thus far giving them a random point here or there is as good a method as the polls for guessing what they'll have on election day.

    Auckland • Since Nov 2006 • 4335 posts Report Reply

  • Hard News: Where are all the polls at?,

    The Herald now has it's modelling up and running.

    As they say by combining all the polls and data from previous elections to correct for poll vs vote inaccuracies they should be able to get a better prediction.

    A nice feature is they actually talk about the limitations, however they fail to mention one of the biggest limitations of such a model which is that if public opinion actually does change significantly then their model will effectively minimise the the influence of such a change. Hence their model has Labour lower and National higher than recent polls predict, if those polls are real, a big if, then the model can't respond to them.

    That said it's really nice to see real variability represented properly rather than the usual "margin of error" bollocks.

    Auckland • Since Nov 2006 • 4335 posts Report Reply

  • Hard News: Where are all the polls at?, in reply to Ian Dalziel,

    they really got creative defining 'Chch Central' electorate last time they tinkered,

    Fortunately with MMP the effect of that is moot for the parliament as a whole, but it can still piss of locals who feel they aren't represented correctly at the electorate level.

    Auckland • Since Nov 2006 • 4335 posts Report Reply

Last ←Newer Page 1 2 3 4 5 434 Older→ First