Posted by John Bibby, QED of York England on April 10, 2015
Why Buy a Quincunx?
John Bibby, QED of York, York, England (qed@enterprise.net)
As one who has actually TRIED building a quincunx, I can only say
it's hellishly difficult: my contraption looked quincunxy, but the
distributions produced were far from normal.
I am left concluding (a) I'm frightfully cack-handed, or (b) Galton
had a fine crafts-team (probably both).
Incidentally, Galton's original quincunx was in the Pearson Building
at University College, London in the 1960's, but now it has gone.
Does anyone know where it is now?
Steven Stigler gives a marvellous discussion if Galton's Q in his
"History of Statistics". This includes a fine demonstration
of how conditional arguments can be used to derive complex results
such as "A normal mixture of normal distributions is itself normal".
Galton is also discussed in Theodore Porter's "The Rise of
Statistical Thinking".