Statistics Good, Bad & Ugly

Lee van Cleef doesn't like your models.
Lee van Cleef doesn’t like your models.

In an attempt to catch up on my 300-some emails (yes, the total has grown considerably; probably because of recent publicity), here are some articles sent in by readers that bear attention.

Wee p-value surge

From our friend John Cook, the paper (pdf) “A surge of p-values between 0.040 and 0.049 in recent decades (but negative results are increasing rapidly too)”.

Whatever the theory is, the result is that P-values are magical thinking.

Replication

Reader Al Perrella sends in “Why Psychologists’ Food Fight Matters: Important findings haven’t been replicated, and science may have to change its ways.

More wee p-values, but in disguise. The real reason for the “replication” or “reproducibility crisis” is revealed here, in The True Meaning Of Statistical Models.

Only when users of statistical models think of them, as physicists think of theirs, in a predictive sense, and thus become verifiable, will the crisis dissipate.

Placebos

Perrella also sent “Placebos work — even without deception”. How does knowing you’re receiving a placebo and either getting better or not differ from not getting a known-placebo?

The headline is busted in the usual way. It makes it sound like known-placebos always work, even though in the quoted experiment it is clearly seen that they don’t.

People who beat each other have better health

Reader I. Fox writes, “Recently, BDSM practitioners have put out studies that say they have better mental health than those overall, and that, like homosexuality, is a perfectly normal behaviour. This document, written by a queer child psychiatrist (conflict of interest is noted) with the usual emotional arguments.”

The paper (pdf) is “Psychology & BDSM: Pathology or Individual Difference?” by Margaret Nichols, which opens “As a clinical psychologist, I am a member of a profession that many believe has replaced religion in its power to influence social opinion and behavior.”

And off she goes trying to influence. She calls her pals, “the kinky community.” Her “work” naturally excited the minds of those who contribute to Live Science: Bondage Benefits: BDSM Practitioners Healthier Than ‘Vanilla’ People. “[S]ome psychiatrists see the inclusion of BDSM and other kinks in the manual as stigmatizing”.

Heaven forfend perversion should be “stigmatizing.” Equality will be our death.

Don’t think so? Fox also sent this: Trans-Uterus, in which men pretending to be women are given uteruses (uterii?) and who then pretend they might get pregnant.

Which is fine. Hey, who am I to judge? But it’s not fine when you insist I pretend too. That’s tyranny.

Book recommendation

Reader Chris writes,

I’ve been reading Standard Deviations by Gary Smith, and think of your blog every time I turn a page. It’s a popular work for sure, and not very heavy on philosophy. I’m not sure if prof. Smith is a logical probabilist, or what, however he touches on so, so many topics you’ve covered over the years. Regression to the mean, the “law of averages”, Texas sharp-shooter, correlation/causation, file-drawer effect and over-certainty in general.

Among the many studies he thoroughly debunks are the “abortion leads to crime reduction”, “successful businesses become mediocre and that’s what keeps our economy running”, and “EMFs from power-lines cause cancer in children”.

You could probably finish it in a day or two. I wanted to send the recommendation your way in case you’re ever lonely and feeling quixotic about fighting this uphill battle against statistical over-certainty.

I haven’t seen it yet, but looks like it could be good.

Antibiotics linked to child obesity

The beauty of the phrase “linked to” is that it means anything you want it do. Thanks to reader Alan Watt for alerting us to the article “Children who receive a lot of antibiotics before age 2 are slightly more likely than others to become obese, a new study shows.

Slightly. Add in the model uncertainty and that due to concentrating on wee p-values and parameters and not observables and make a guess what will happen.

Highway Help!

Reader Jason asks us:

My wife, child, and myself are considering a move to a property that is approximately 860 feet from HWY 5 in San Juan Capistrano. My wife is pregnant and we have been reading about the harmful effects of living too close to a major highway.

My question is if you would consider this distance from the highway exceptionally unsafe?

Only if you play in the traffic.

Bayes

More Bayes in the news, sent in by reader John B.

Harry Potter

Our friend Ye Olde Statistician alerts us to the breathless statistical study, “D​id Harry Potter Influence The Political Views of Millennials?

The right answer: probably not, but who knows?

Survey says

YOS also shows us the absurdity of surveys: “An item on the Beeb alerted me to the fact that the Danes have — yet again — scored highest in some international measurement of happiness levels.

9 Comments

  1. Scotian

    This quote “Margaret Nichols, which opens “As a clinical psychologist, I am a member of a profession that many believe has replaced religion in its power to influence social opinion and behavior.”” reminds me of this recent blog:

    http://stuartschneiderman.blogspot.ca/2014/10/is-psychoanalysis-scam.html

    and this line within “Of course, Lacan was not telling his followers to toss their Freud books in the poubelle. He wanted psychoanalysis to fulfill its destiny by becoming an instrument of cultural revolution and thought reform. He wanted to lead psychoanalysis out of the clinic and into the cultural arena.”

  2. Sheri

    Jason: It would be to me–can’t stand living that close to anything. However, as far as health goes, I wouldn’t worry. Interestingly, one of my past acquaintances moved away from town to the “healthy” country and was described as “looking awful” by his neighbor. Not sure there’s any causality in any of this, of course, but anecdotes run both ways.

    Concerning Harry Potter: I thought “more open to diversity and are more politically tolerant” described most college students. I am surprised the study found any link to what the students had read.

    On the Dane article: I just love this line “Unfortunately, the Internet has now pulled out all the stops, and people can scare themselves without organized assistance.”

  3. MattS

    Briggs,

    On Replication:

    “Only when users of statistical models think of them, as physicists think of theirs, in a predictive sense, and thus become verifiable, will the crisis dissipate. ”

    Actually, the problems with replication extend to physics as well, the problem just isn’t a pronounced there.

    There is more to this problem than just wrong thinking about models. Even if the wrong thinking about models is corrected, the problem will persist until the following two issues are also corrected.

    1. Journals don’t like to publish replication studies, particularly replication studies with negative results.

    2. Universities don’t want their faculty working on replication studies, as these tend not to bring in grant money.

  4. MattS

    “Fox also sent this: Trans-Uterus, in which men pretending to be women are given uteruses (uterii?) and who then pretend they might get pregnant. ”

    Actually, if a trans-gender uterine transplant works, is there any reason why they wouldn’t be able to transplant ovaries as well? If a trans gender uterus and ovary transplant works, would they still be pretending?

  5. Don Jackson

    From the preprint of the “a surge of p values” paper:

    […} Fanelli found that the number of papers providing support for the main hypothesis had increased from 70% in 1990 to 86% in 2007 (it is unclear why Fanelli reported an over 22% increase in the abstract).

    Is it too much to ask, that a researcher be numerate enough to recognize 16/70? 🙂

  6. Sheri

    Matt S: Assuming Mr. Mom actually is possible, no, they are not pretending they can become pregnant. They are, however, still pretending to be “female”.

  7. John B

    …and I got it “More Bayes in the News” from Bishop Hil…

    MattS:

    How did you ever transition from replication to replication

    Sheri:

    Mr Mom? – why not the Rabbit Test?

  8. Sheri

    John B: I had not ever heard of the Rabbit Test movie. Actually, I should have gone with “Junior”. (Mr. Mom just was the first thing I thought of.)

  9. Bruce Foutch

    The trans-uterus conversation made me think of this little life story. No idea why other than it seemed to fit:

    A little boy and little girl were talking and the little boy spouted that “Boys are better than girls.”

    “No way!” said the little girl.

    “Yes we are.” replied the little boy, who proceeded to unzip his pants and point to his little boyhood. “Boys are better because we have one of these.”

    “Oh, no way!” said the little girl, who lifted up her dress and pointed at her little girlhood. “Girls have one of these! And, with one of these we can get all of those we want!”

Leave a Reply

Your email address will not be published. Required fields are marked *