Group differences

« previous post | next post »

And not just social scientists…

Some background:

"Gabby guys: the effect size", 9/23/2006
"The Pirahã and us", 10/6/2007
"Pop platonism and unrepresentative samples", 7/26/2008
"Mandatory treatment for generic plurals", 9/13/2009
"Generic comparisons", 11/7/2011
"Generic plural nurses", 9/22/2018

Of course, there are some real group differences (see e.g. "Biology, sex, culture, and pitch", 8/16/2013;"Age, sex, and f0", 3/25/2017); but the fraction of the American population that thinks clearly about these issues is roughly the same as the fraction of the Pirahã population skilled in mental arithmetic.



16 Comments

  1. Antonio L. Banderas said,

    November 18, 2018 @ 1:45 pm

    Could sb please elaborate a bit more? Thanks

  2. David Marjanović said,

    November 18, 2018 @ 3:19 pm

    And not just of the American population…!

  3. D.O. said,

    November 18, 2018 @ 4:35 pm

    It's telling that even a picture that skewers public misunderstanding of statistics, shows nice symmetric, short tail distributions…

  4. KDMaus said,

    November 18, 2018 @ 6:02 pm

    As far as caricatures of what the ill-informed plebs think, it seems more apt to say: 'people think' it's either perfect overlap or no overlap, insofar as people think about it at all

    And then the caricature of what polemically-inclined specialists think would be that "the other side grossly underestimates/overestimates the differences or their relevance"

    And then the caricature of all us rational folk, not doing rhetoric and not ill-informed, just tell it like it is?

    On this division, the person posting what 'social scientists think' likely falls into the second polemically-inclined category. Example of polemic: "Those people want to say the group differences don't matter/exist, they're so wrong!" or "Those people wants to say the group differences are huge/all that matter, they're so wrong!" ….And when you force two polemical specialists into the same conversation, it just slowly deflates into who wants to emphasize what and why, or it meanders into pettiness

  5. eub said,

    November 19, 2018 @ 12:40 am

    Cohen (of the d) proposed to standardize terms like "small difference", "large difference" based on ranges of effect size. Is there any reason why this would not be miles better than current science discourse, or was it just hard to convince enough people to care?

    [(myl) Many people, including me, spend a lot of effort trying to get scientists (and science writers) to use (things like) effect sizes rather than p values, and to talk about practical rather than statistical significance. There are some cultural changes in the right direction.]

  6. John Swindle said,

    November 19, 2018 @ 4:26 am

    @Antonio L. Banderas
    I think it's specifically about grouping humans. Imagine that the red and blue in the chart represent, say, men and women. Compare them on height, weight, politics, musical preference, or love for insects. Are the groups completely different? That's the figure at right. Are average differences dwarfed by differences with the groups? That's the figure at left. People of different groups turn out to be a lot alike.

  7. Bill Benzon said,

    November 19, 2018 @ 5:03 am

    In a reply Geoffrey Miller asked him to "do it with actual data." He replied with a graph he'd done based on data from a study recently published in Science: https://twitter.com/vaiseys/status/1053062916356890624

  8. J.W. Brewer said,

    November 19, 2018 @ 11:28 am

    One bonus here is that both of the rival options have unlabeled x and y axes, making it impossible to gauge how reasonable/common-sensical the actual shapes of the resulting curves are as a presentation of the underlying data.

  9. Meredith said,

    November 19, 2018 @ 2:22 pm

    As a statistician, this is one of the most difficult things to get people (including researchers) to understand. MYL – I'm all in favor of moving away from p-values, especially to confidence intervals since they clearly show how large the within-group variability is compared to that between-groups. The hardest part is convincing certain journals to allow that move.

  10. Stan Carey said,

    November 19, 2018 @ 2:30 pm

    Der Narzissmus der kleinen Differenzen.

  11. Andrew Usher said,

    November 19, 2018 @ 6:02 pm

    Re Bill Benzon:

    That study quoted there in your link was about differences of differences, and so the absolute differences should be much greater. Not that the statistical point is wrong – if, that is, all variables were normally distributed!

    I also noted glancing at that paper's page that the 'conclusion' started with a comment that says the study supports the need for 'gender equality'. Can anyone doubt that they'd have started exactly the same way had they found the opposite? In social sciences, and sometimes other sciences, conclusions can be worthless even if the data are unimpeachable.

    k_over_hbarc at yahoo dot com

  12. Rick Rubenstein said,

    November 20, 2018 @ 6:51 pm

    This stems from the fact that Westerners see the world as consisting of dichotomies, whereas Asians think holistically. [Imagine this space is occupied by a winking emoticon of transcendent power.]

  13. Philip Anderson said,

    November 21, 2018 @ 8:49 am

    Rick Rubenstein:
    That’s not a fact, it’s a dichotomy :-)

    It’s a generalisation and an ambiguous one. In the UK, Asian means from an ethnic group originating in South Asia, the Indian subcontinent, but I guess you mean East Asian?

  14. Xenobio said,

    November 21, 2018 @ 4:13 pm

    Some wag commented on the fury over the IARC ruling that preserved meats are carcinogenic: "Bacon: Teaching people the difference between statistical significance and effect size."

  15. The Other Mark P said,

    November 21, 2018 @ 5:07 pm

    Except, of course, that we are all used to seeing group that are much closer to the one on the right. If there are group differences as tight as the one on the left, we would barely even call it in normal life. People may not understand standard deviations and the like, but they are not idiots. They know that there are substantial group differences.

    Arm strength of males and females barely overlap — the weakest man is a match for all but the strongest woman. The amount of facial hair of Europeans vs the amount of facial hair of Japanese — most Europeans can grow a full beard and few Japanese can come close.

    And there are important real world differences between groups that anyone would be a fool to ignore. The maturity of sixteen year olds vs the maturity of six year olds is not a small overlap, but exactly like the picture on the right. We do, and should, treat six year olds and sixteen year olds differently because they are massively different. We do that, even though we know six and sixteen year olds themselves have some variation in maturity. Only a moron would suggest that teachers treat Grade 1 and Grade 11 the same because their is internal differences inside each group.

    There's a campaign to try and suppress anything that smacks of group differences. Don't be fooled. Most differences between sexes, races, sexual orientations etc
    are like the overlap on the left — where internal variation overwhelms group difference. But not all differences are of that sort.

    In particular, we know that the curve for male intelligence and that for female intelligence are not the same shape. They have the same averages, but they do not have similar standard deviations. We've known this for a long time, and it used to be non-controversial. Now it is forbidden knowledge — or rather the consequences of it are.

    People put out cutesy explanations of how things "really look" that simply are not true in every case.

  16. Philip Anderson said,

    November 21, 2018 @ 5:50 pm

    @The Other Mark P
    “Most Europeans can grow a full beard”. Really? How many European women have you met with beards?

    I am a little sceptical about your arm-wrestling claim too.

RSS feed for comments on this post