Sunday, May 19, 2019

The Statistician's Error?

I just attended (and gave a talk at) the United States Conference on Teaching Statistics (USCOTS). Big thanks to Allan Rossman, who brilliantly MC-ed the conference.

• One keynote was about "moving beyond p < .05" in a talk by Ron Wasserstein and Allen Schirm, In their recent editorial in The American Statistician (with Nicole Lazar), a primary recommendation was Don't Say "Statistically Significant". Decisions with p values are about controlling error rates, but dichotomous decisions let people slip into "bright line" thinking wherein p < .05 means real and important and p > .05 means absent and unimportant.

• Another keynote, in a talk by Kari Lock Morgan, was about three possible explanations of an apparent effect of a manipulation, namely (i) genuine cause, (ii) random difference at baseline before manipulation, and (iii) random difference after manipulation.

I returned home from the conference this morning. To relax, after the intensive pre-conference preparation and during-conference insomnia, I opened a book of poetry and came across a poem by Aaron Fogel that (inadvertently) reflects upon both talks. It's a poem about how editors of printing make decisions regarding errors, and about three sources or errors, and distinguishing the sources of error. And about the role of editors (and perhaps of statisticians?).


The Printer's Error
Fellow compositors
and pressworkers!
I, Chief Printer
Frank Steinman,
having worked fifty-
seven years at my trade,
and served five years
as president
of the Holliston
Printer's Council,
being of sound mind
though near death,
leave this testimonial
concerning the nature
of printers' errors.
First: I hold that all books
and all printed
matter have
errors, obvious or no,
and that these are their
most significant moments,
not to be tampered with
by the vanity and folly
of ignorant, academic
textual editors.
Second: I hold that there are
three types of errors, in ascending
order of importance:
One: chance errors
of the printer's trembling hand
not to be corrected incautiously
by foolish professors
and other such rabble
because trembling is part
of divine creation itself.
Two: silent, cool sabotage
by the printer,
the manual laborer
whose protests
have at times taken this
historical form,
covert interferences
not to be corrected
censoriously by the hand
of the second and far
more ignorant saboteur,
the textual editor.
Three: errors
from the touch of God,
divine and often
obscure corrections
of whole books by
nearly unnoticed changes
of single letters
sometimes meaningful but
about which the less said
by preemptive commentary
the better.
Third: I hold that all three
sorts of error,
errors by chance,
errors by workers' protest,
and errors by
God's touch,
are in practice the
same and indistinguishable.
Therefore I,
Frank Steinman,
typographer
for thirty-seven years,
and cooperative Master
of the Holliston Guild
eight years,
being of sound mind and body
though near death
urge the abolition
of all editorial work
whatsoever
and manumission
from all textual editing
to leave what was
as it was, and
as it became,
except insofar as editing
is itself an error, and
therefore also divine.

Wednesday, March 13, 2019

Bayesian statistics at Princeton University, with a visit to the grave of John Von Neumann

I was very pleased to have had the opportunity to present a talk regarding Bayesian models of ordinal data at Princeton University, on Tuesday, March 5, 2019. An abstract of the talk is here, and the published article on which the talk was based is here.

A big thanks to Dr. Ting Qian who orchestrated the visit wonderfully. The lecture hall was packed, with people standing outside the door in the hall, in no small part due to Ting's organization of a popular statistics series. Big thanks also to the staff who made arrangements. And, of course, thanks to the many people who took time to meet with me while I was there.

I took a few hours the next morning to visit some special places in Princeton. In particular, I visited the graves of John Von Neumann and Kurt Godel. Von Neumann made contributions to Monte Carlo methods at the foundation of MCMC methods in Bayesian analysis.

Here's a snapshot of Von Neumann's grave stone:

And only a few feet away, Godel's grave stone:

After visiting the cemetery, I went to the Princeton University Art Museum and came across this painting by Georgia O'Keeffe. It's titled, "From a New Jersey Weekend II," painted in 1941.
It was an interesting coincidence that both O'Keeffe's and my visits featured two notable grave stones.

Here's a map of the Princeton Cemetery, with a marker at the approximate location of the graves of Von Neumann and Godel.


For posts about the book, DBDA2E, visiting other famous grave sites, see this post and its links to previous posts.

Thursday, February 21, 2019

A Stendhal moment on the way to Bayesian stats class

On the way to my Bayesian stats class this morning I had a few moments of Stendhal syndrome and thought I'd share it with y'all. (Aside from being on the way to Bayesian stats class, it has nothing directly relevant to Bayesian statistics.) The sun was shining through some construction fences lining a sidewalk and were unintentionally beautifying an otherwise routine Thursday morning. Here are a couple of snapshots from my office window:


It reminds me of Christo and Jeanne-Claude's Running Fence: