Log in

Unfortunate characterisation of the normal distribution

  • 19 Feb 2022 8:00 AM
    Reply # 12610961 on 12608364
    Jessica Kasza wrote:

    ". . . I think an issue that the author of the piece in the Scientific American is trying to highlight with the statement is that a single distribution or statistical model cannot describe all."

    The normal distribution is, of course, "just" another model.  All we have to make sense of the outside  world are models.  In everyday life, these are the models that our brain forms  from the sensory inputs that are available to it.  Mathematical models are, in  this context, another part of the toolkit.

    Models always operate in a specific context.  Their use in this context requires very careful critique, a point that it is hardly possible to over-emphasize in statistics teaching.

    One continues to see claims that 75% of variation in IQ, as measured by IQ tests, is genetic.  At one point in time, in a specific population, assuming that the effects of environment and heredity can be separated, maybe. In a wider context, it does not account for the Flynn effect by which IQ scores increased, in some places over a period of 60 or 70 years, by more than a standard deviation. 

    All, no doubt, too subtle for the author of the Scientific American article.

  • 18 Feb 2022 7:38 AM
    Reply # 12608364 on 12605808
    Jessica Kasza (Administrator)

    I'd question whether Scientific American was ever thought of as a reputable journal - I think it's always been more of a magazine style of publication! 

    In any case, I would agree that the characterisation of the normal distribution in the magazine is unfortunate indeed. I won't comment on the rest of the content of that article, but I think an issue that the author of the piece in the Scientific American is trying to highlight with the statement is that a single distribution or statistical model cannot describe all.  As statisticians we are well aware of this - it's why we include covariates in our regression models!

  • 17 Feb 2022 4:40 PM
    Reply # 12606601 on 12605808

    Hi Chris

    You have drawn me into a post with the link to that article. I have decided to try not to say too much at all.

    I agree that the description of the normal distribution is unfortunate in a scientific paper or a pop science magazine like Scientific American. It detracts from any serious discussion that would be needed to cover all the issues raised.

    I could critique that article forever but am not going to. I also am concerned at the level of some previously mid-quality pop science journals. I have already raised those concerns elsewhere

    However the issues in the article are immense. Just the article detracts somewhat from even discussing them well or safely

    Many of us have our own personal struggles with how science and lack of knowledge of methods are used, I hate to say abused, on a daily basis

    As for mixing science and politics was it not always such. And I believe from looking at your online site some of your interest areas are in social sciences. Are they not

    This is the final version of my post. I promise

    But in summary Chris, you drew me into a post and I agree with the word "unfortunate" and I would love to be a first year student given that as an assignment and the freedom to critique. Not the paper. The article

    regards Duncan

    Last modified: 18 Feb 2022 9:04 AM | Duncan Lowes
  • 17 Feb 2022 9:31 AM
    Reply # 12605934 on 12605808

    Great suggestion, Chris.  

  • 17 Feb 2022 8:39 AM
    Message # 12605808

    Scientific American was once a reputable journal. E.O. Wilson is only dead a couple of months and they publishes this pejorative assessment by an associate professor of Family Health Care. Why should you care? Here is a quote that should raise the interest of the society:

    "The so-called normal distribution of statistics assumes that there are default humans who serve as the standard that the rest of us can be accurately measured against." The default humans term links to a paper in a C journal which looks like it might have been written by a bot. When you mix science and politics you get … politics.

    How about the Society write to Scientific American stating clearly that her characterisation of the normal distribution is not to be taken seriously?

Powered by Wild Apricot Membership Software