A Pearson Person

All my post-grad life I've been buying up 25c copies of MJ Moroney's book Facts From Figures (1951) and giving them to younger people who have numbers to crunch. Because statistics is a) essential for effective science b) usually taught so poorly across these islands. Moroney strikes an agreeable balance between correct and accessible, but it's dated by being written before calculators, let alone Excel SPSS and R. As a 12 y.o. could calculate the variance of a [small] dataset with the help of a pencil, squared paper and Charlier's Checks. Nobody would consider teaching the skill to Gdau.I at the same age.

I was in Wexford Town Library early before Science Cafe in mid-December and found a copy of David Spiegelhalter's The Art of Statistics, Learning from Data (2020) call out to me from the 519.5 shelving. Spiegelhalter is frequently wheeled on stage for the Tim Harford's More of Less podcast when some egregious numerical bloomer needs to be exposed and explained to Joe Public . . . without being either stodgy, geeky or patronising.

There are some arresting patriarch-involving images. In explaining the prosecutor's fallacy, Spiegelhalter uses the following as a reductio ad absurdem: if you're the pope, then you're a Catholic is not the same as if you're a Catholic then you're the pope. Later, in a discussion of Bayesian statistics, he considers likelihood ratio = (the prob of dealing himself a royal flush, assuming the Archbishop of Canterbury is cheating) ÷ (the prob of royal flush assuming the ABofC is lucky). Which was funnier because absurd when the sentence was written than now with the 2024 Archbishop of Canterbury sheltering abusive paedophiles and resigning only with reluctance.

The self-styled Chevalier de Méré was a 17thC rake who gambled A Lot. He wanted to know which of two dicey games of chance was mostly likely to make him long-term money: a) throw a single dice 4x to get a single ⚅ or b) throw pairs of dice 24x to land ⚅⚅ [wrong wrong almost R]. Neither he, nor anybody else at the time, knew the correct mathematical way to calculate the odds, so he set to and rolled a heckuva lot of dice to [correctly] determine the answer. Hanc marginis exiguitas non caperet This is exactly what I did to determine the likelihood of winning while playing Klondike patience. The Chev wanted the right answer mathematically to save other gamblers from doing a similar multiple trial experiment and presented it to Blaise "God's Wager" Pascal [prev] who in turn shared it with Pierre "Hanc marginis exiguitas non caperet" de Fermat [prev]. Between them they started the modern ride in probability theory. People like me, who started calculating means standard deviations on paper ~60 years ago are now trying to get their heads around Bayesian stats and  prior and posterior probabilities. Spiegelhalter 2020 is helpful in this regard, while Moroney1950  ignored the issue entirely.

Reveal. Pearson Person? Both Karl "correlation" Pearson and son Egon "confidence" Pearson get a shout. But really, at heart, Spiegelhalter is a Bayesian Bloke