## limited shelf validity

Posted in Books, pictures, Statistics, University life with tags , , , , , , , , , , , on December 11, 2019 by xi'an

A great article from Steve Stigler in the new, multi-scaled, and so exciting Harvard Data Science Review magisterially operated by Xiao-Li Meng, on the limitations of old datasets. Illustrated by three famous datasets used by three equally famous statisticians, Quetelet, Bortkiewicz, and Gosset. None of whom were fundamentally interested in the data for their own sake. First, Quetelet’s data was (wrongly) reconstructed and missed the opportunity to beat Galton at discovering correlation. Second, Bortkiewicz went looking (or even cherry-picking!) for these rare events in yearly tables of mortality minutely divided between causes such as military horse kicks. The third dataset is not Guinness‘, but a test between two sleeping pills, operated rather crudely over inmates from a psychiatric institution in Kalamazoo, with further mishandling by Gosset himself. Manipulations that turn the data into dead data, as Steve put it. (And illustrates with the above skull collection picture. As well as warning against attempts at resuscitating dead data into what could be called “zombie data”.)

“Successful resurrection is only slightly more common than in Christian theology.”

His global perspective on dead data is that they should stop being used before extending their (shelf) life, rather than turning into benchmarks recycled over and over as a proof of concept. If only (my two cents) because it leads to calibrate (and choose) methods doing well over these benchmarks. Another example that could have been added to the skulls above is the Galaxy Velocity Dataset that makes frequent appearances in works estimating Gaussian mixtures. Which Radford Neal signaled at the 2001 ICMS workshop on mixture estimation as an inappropriate use of the dataset since astrophysical arguments weighted against a mixture modelling.

“…the role of context in shaping data selection and form—context in temporal, political, and social as well as scientific terms—has been shown to be a powerful and interesting phenomenon.”

The potential for “dead-er” data (my neologism!) increases with the epoch in that the careful sleuth work Steve (and others) conducted about these historical datasets is absolutely impossible with the current massive data sets. Massive and proprietary. And presumably discarded once the associated neural net is designed and sold. Letting the burden of unmasking the potential (or highly probable?) biases to others. Most interestingly, this recoups a “comment” in Nature of 17 October by Sabina Leonelli on the transformation of data from a national treasure to a commodity which “ownership can confer and signal power”. But her call for openness and governance of research data seems as illusory as other attempts to sever the GAFAs from their extra-territorial privileges…

Posted in Books, Kids, R with tags , , , , , , , on November 19, 2019 by xi'an

Since Galton’s quincunx has fascinated me since the (early) days when I saw a model of it as a teenager in an industry museum near Birmingham, I jumped on the challenge to build an uneven nail version where the probabilities to end up in one of the boxes were not the Binomial ones. For instance,  producing a uniform distribution with the maximum number of nails with probability ½ to turn right. And I obviously chose to try simulated annealing to figure out the probabilities, facing as usual the unpleasant task of setting the objective function, calibrating the moves and the temperature schedule. Plus, less usually, a choice of the space where the optimisation takes place, i.e., deciding on a common denominator for the (rational) probabilities. Should it be 2⁸?! Or more (since the solution with two levels also involves 1/3)? Using the functions

evol<-function(P){
Q=matrix(0,7,8)
Q[1,1]=P[1,1];Q[1,2]=1-P[1,1]
for (i in 2:7){
Q[i,1]=Q[i-1,1]*P[i,1]
for (j in 2:i)
Q[i,j]=Q[i-1,j-1]*(1-P[i,j-1])+Q[i-1,j]*P[i,j]
Q[i,i+1]=Q[i-1,i]*(1-P[i,i])
Q[i,]=Q[i,]/sum(Q[i,])}
return(Q)}


and

temper<-function(T=1e3){
bestar=tarP=targ(P<-matrix(1/2,7,7))
temp=.01
while (sum(abs(8*evol(R.01){
for (i in 2:7)
R[i,sample(rep(1:i,2),1)]=sample(0:deno,1)/deno
if (log(runif(1))/temp<tarP-(tarR<-targ(R))){P=R;tarP=tarR}
for (i in 2:7) R[i,1:i]=(P[i,1:i]+P[i,i:1])/2
if (log(runif(1))/temp<tarP-(tarR<-targ(R))){P=R;tarP=tarR}
if (runif(1)<1e-4) temp=temp+log(T)/T}
return(P)}


I first tried running my simulated annealing code with a target function like

targ<-function(P)(1+.1*sum(!(2*P==1)))*sum(abs(8*evol(P)[7,]-1))

where P is the 7×7 lower triangular matrix of nail probabilities, all with a 2⁸ denominator, reaching

60
126 35
107 81 20
104 71 22 0
126 44 26 69 14
61 123 113 92 91 38
109 60 7 19 44 74 50

for 128P. With  four entries close to 64, i.e. ½’s. Reducing the denominator to 16 produced once

8
12 1
13 11 3
16  7  6   2
14 13 16 15 0
15  15  2  7   7  4
8   0    8   9   8  16  8

as 16P, with five ½’s (8). But none of the solutions had exactly a uniform probability of 1/8 to reach all endpoints. Success (with exact 1/8’s and a denominator of 4) was met with the new target

(1+,1*sum(!(2*P==1)))*(.01+sum(!(8*evol(P)[7,]==1)))

imposing precisely 1/8 on the final line. With a solution with 11 ½’s

0.5
1.0 0.0
1.0 0.0 0.0
1.0 0.5 1.0 0.5
0.5 0.5 1.0 0.0 0.0
1.0 0.0 0.5 0.0 0.5 0.0
0.5 0.5 0.5 1.0 1.0 1.0 0.5

and another one with 12 ½’s:

0.5
1.0 0.0
1.0 .375 0.0
1.0 1.0 .625 0.5
0.5  0.5  0.5  0.5  0.0
1.0  0.0  0.5  0.5  0.0  0.5
0.5  1.0  0.5  0.0  1.0  0.5  0.0

Incidentally, Michael Proschan and my good friend Jeff Rosenthal have an 2009 American Statistician paper on another modification of the quincunx they call the uncunx! Playing a wee bit further with the annealing, and using a denominator of 840 let to a 60P  with 13 ½’s out of 28

30
60 0
60 1 0
30 30 30 0
30 30 30 30 30
60  60  60  0  60  0
60  30  0  30  30 60 30

## down with Galton (and Pearson and Fisher…)

Posted in Books, Statistics, University life with tags , , , , , , , , , , , , , , , on July 22, 2019 by xi'an

In the last issue of Significance, which I read in Warwick prior to the conference, there is a most interesting article on Galton’s eugenics, his heritage at University College London (UCL), and the overall trouble with honouring prominent figures of the past with memorials like named building or lectures… The starting point of this debate is a protest from some UCL students and faculty about UCL having a lecture room named after the late Francis Galton who was a professor there. Who further donated at his death most of his fortune to the university towards creating a professorship in eugenics. The protests are about Galton’s involvement in the eugenics movement of the late 18th and early 19th century. As well as professing racist opinions.

My first reaction after reading about these protests was why not?! Named places or lectures, as well as statues and other memorials, have a limited utility, especially when the named person is long dead and they certainly do not contribute in making a scientific theory [associated with the said individual] more appealing or more valid. And since “humans are [only] humans”, to quote Stephen Stigler speaking in this article, it is unrealistic to expect great scientists to be perfect, the more if one multiplies the codes for ethical or acceptable behaviours across ages and cultures. It is also more rational to use amphitheater MS.02 and lecture room AC.18 rather than associate them with one name chosen out of many alumni’s or former professors’.

Predictably, another reaction of mine was why bother?!, as removing Galton’s name from the items it is attached to is highly unlikely to change current views on eugenism or racism. On the opposite, it seems to detract from opposing the present versions of these ideologies. As some recent proposals linking genes and some form of academic success. Another of my (multiple) reactions was that as stated in the article these views of Galton’s reflected upon the views and prejudices of the time, when the notions of races and inequalities between races (as well as genders and social classes) were almost universally accepted, including in scientific publications like the proceedings of the Royal Society and Nature. When Karl Pearson launched the Annals of Eugenics in 1925 (after he started Biometrika) with the very purpose of establishing a scientific basis for eugenics. (An editorship that Ronald Fisher would later take over, along with his views on the differences between races, believing that “human groups differ profoundly in their innate capacity for intellectual and emotional development”.) Starting from these prejudiced views, Galton set up a scientific and statistical approach to support them, by accumulating data and possibly modifying some of these views. But without much empathy for the consequences, as shown in this terrible quote I found when looking for more material:

“I should feel but little compassion if I saw all the Damaras in the hand of a slave-owner, for they could hardly become more wretched than they are now…”

As it happens, my first exposure to Galton was in my first probability course at ENSAE when a terrific professor was peppering his lectures with historical anecdotes and used to mention Galton’s data-gathering trip to Namibia, literally measure local inhabitants towards his physiognomical views , also reflected in the above attempt of his to superpose photographs to achieve the “ideal” thief…

## a quincunx on NBC

Posted in Books, Kids, pictures, Statistics with tags , , , , , , , , , , on December 3, 2017 by xi'an

Through Five-Thirty-Eight, I became aware of a TV game call The Wall [so appropriate for Trumpian times!] that is essentially based on Galton’s quincunx! A huge [15m!] high version of Galton’s quincunx, with seven possible starting positions instead of one, which kills the whole point of the apparatus which is to demonstrate by simulation the proximity of the Binomial distribution to the limiting Normal (density) curve.

But the TV game has obvious no interest in the CLT, or in the Beta binomial posterior, only in a visible sequence of binary events that turn out increasing or decreasing the money “earned” by the player, the highest sums being unsurprisingly less likely. The only decision made by the player is to pick one of the seven starting points (meaning the outcome should behave like a weighted sum of seven Normals with drifted means depending on the probabilities of choosing these starting points). I found one blog entry analysing an “idiot” strategy of playing the game, but not the entire game. (Except for this entry on the older Plinko.) And Five-Thirty-Eight surprisingly does not get into the optimal strategies to play this game (maybe because there is none!). Five-Thirty-Eight also reproduces the apocryphal quote of Laplace not requiring this [God] hypothesis.

[Note: When looking for a picture of the Quincunx, I also found this desktop version! Which “allows you to visualize the order embedded in the chaos of randomness”, nothing less. And has even obtain a patent for this “visual aid that demonstrates [sic] a random walk and generates [re-sic] a bell curve distribution”…]

## another Galton-Watson riddle

Posted in Statistics with tags , , , on February 3, 2017 by xi'an

The riddle on the Riddler this week is definitely a classic, since it rephrases the standard Galton-Watson branching process (which should have been called Bienaymé‘s process, as he established the relation before Watson, while the jack-of-all-trades Francis Galton only posed the question):

At the beginning, there is a single microorganism. Each day, every member of this species either splits into two copies of itself or dies. If the probability of multiplication is p, what are the chances that this species goes extinct?

As is easily seen from the moment generating function, the species goes instinct if p≤½. Actually, I always found it intriguing [intuitively] that the value ½ is included in the exclusion range!

## a Galton-Watson riddle

Posted in R, Travel with tags , , , on December 30, 2016 by xi'an

The Riddler of this week has an extinction riddle which summarises as follows:

One observes a population of N individuals, each with a probability of 10⁻⁴ to kill the observer each day. From one day to the next, the population decreases by one individual with probability

K√N 10⁻⁴

What is the value of K that leaves the observer alive with probability ½?

Given the sequence of population sizes N,N¹,N²,…, the probability to remain alive is

$(1-10^{-4})^{N+N^1+\ldots}$

where the sum stops with the (sure) extinction of the population. Which is the moment generating function of the sum. At x=1-10⁻⁴. Hence the problem relates to a Galton-Watson extinction problem. However, given the nature of the extinction process I do not see a way to determine the distribution of the sum, except by simulation. Which returns K=27 for the specific value of N=9.

N=9
K=3*N
M=10^4
vals=rep(0,M)
targ=0
ite=1
while (abs(targ-.5)>.01){

for (t in 1:M){
gen=vals[t]=N
while (gen>0){
gen=gen-(runif(1)<K*sqrt(gen)/10^4)
vals[t]=vals[t]+gen}
}
targ=mean(exp(vals*log(.9999)))
print(c(as.integer(ite),K,targ))
if (targ<.5){ K=K*ite/(1+ite)}else{
K=K/(ite/(1+ite))}
ite=ite+1}


The solution proposed on The Riddler is more elegant in that the fixed point equation is

$\prod_{J=1}^9 \frac{K \cdot \sqrt{J}}{K \cdot \sqrt{J} + J}=\frac{1}{2}$

with a solution around K=27.

## Tractable Fully Bayesian inference via convex optimization and optimal transport theory

Posted in Books, Statistics, University life with tags , , , , , , , , on October 6, 2015 by xi'an

“Recently, El Moselhy et al. proposed a method to construct a map that pushed forward the prior measure to the posterior measure, casting Bayesian inference as an optimal transport problem. Namely, the constructed map transforms a random variable distributed according to the prior into another random variable distributed according to the posterior. This approach is conceptually different from previous methods, including sampling and approximation methods.”

Yesterday, Kim et al. arXived a paper with the above title, linking transport theory with Bayesian inference. Rather strangely, they motivate the transport theory with Galton’s quincunx, when the apparatus is a discrete version of the inverse cdf transform… Of course, in higher dimensions, there is no longer a straightforward transform and the paper shows (or recalls) that there exists a unique solution with positive Jacobian for log-concave posteriors. For instance, log-concave priors and likelihoods. This solution remains however a virtual notion in practice and an approximation is constructed via a (finite) functional polynomial basis. And minimising an empirical version of the Kullback-Leibler distance.

I am somewhat uncertain as to how and why apply such a transform to simulations from the prior (which thus has to be proper). Producing simulations from the posterior certainly is a traditional way to approximate Bayesian inference and this is thus one approach to this simulation. However, the discussion of the advantage of this approach over, say, MCMC, is quite limited. There is no comparison with alternative simulation or non-simulation methods and the computing time for the transport function derivation. And on the impact of the dimension of the parameter space on the computing time. In connection with recent discussions on probabilistic numerics and super-optimal convergence rates, Given that it relies on simulations, I doubt optimal transport can do better than O(√n) rates. One side remark about deriving posterior credible regions from (HPD)  prior credible regions: there is no reason the resulting region is optimal in volume (HPD) given that the transform is non-linear.