Janos Galambos: The Mathematics of the Worst Case
Most of statistics is about the middle — averages, typical values, what usually happens. Janos Galambos spent his career on the opposite end: the highest flood in a century, the largest claim in a portfolio, the single worst observation in a sample. In 1978 he published the book that organised that subject, and engineers who have never heard of him have been building to its conclusions ever since.
Zirc
Galambos was born on 1 September 1940 in Zirc, a small town in the Bakony hills of western Hungary, best known for its Cistercian abbey. He entered Eötvös Loránd University in Budapest in 1958 and took his doctorate there in 1963, aged twenty-three.
His supervisor was Alfréd Rényi. That is the single most important fact about his formation. Rényi was the dominant figure of Hungarian probability, the co-author with Erdős of the random graph model that bears both their names, and the founder of a school whose signature was a certain fearless combinatorial style — probability used as a tool for prising open problems in other fields, especially number theory. Galambos absorbed that method and never put it down.
Budapest, Legon, Ibadan, Philadelphia
He stayed at Eötvös Loránd as an assistant professor in 1964–65. Then he left, and the itinerary is worth reading slowly.
From 1965 to 1969 he lectured at the University of Ghana at Legon. From 1969 to 1970 he lectured at the University of Ibadan in Nigeria. Only in 1970 did he arrive at Temple University in Philadelphia, where he then remained on the faculty until his retirement in 2012 — forty-two years.
Five years teaching mathematics in West Africa in the immediate post-independence period is not a standard entry on a probabilist's CV. It was the era when the new universities of Ghana and Nigeria were staffing up fast, and a young Hungarian with a Rényi doctorate was exactly the sort of person they recruited. Whatever else it was, it was not a career-optimising move.
Extreme Values
The work he is remembered for concerns extreme order statistics — the mathematics of maxima and minima.
The central question is deceptively simple. Take a large sample from some distribution and record only the largest value. As the sample grows, what does the distribution of that maximum look like? The answer, remarkably, is that it converges to one of a very small family of limiting shapes almost regardless of what you started with — a result as surprising in its way as the central limit theorem, and far more useful when the thing you fear is not the average but the outlier.
*The Asymptotic Theory of Extreme Order Statistics*, published by Wiley in 1978, gave that theory its systematic modern treatment. It was translated into Russian in 1984 and appeared in a Chinese edition in 2001 — a rough but reliable index of how far a mathematical text has travelled. Galambos is remembered as having significantly shaped how the mathematical community approaches the subject, and he extended it in directions of direct practical use, including nonparametric tests for extreme value distributions.
With H. A. David he wrote a seminal 1974 paper on concomitants of order statistics — what happens to a second, correlated measurement when you select on the first. Anyone who has ever wondered why the tallest man in a room is not reliably the heaviest is asking a version of that question.
Probability as a Crowbar
The Rényi inheritance shows most clearly in Galambos's other line of work: using probability to attack problems that are not about chance at all.
*Representations of Real Numbers by Infinite Series* (Springer, 1976) treats the digits of a number's expansion as if they were random variables and asks what almost all numbers look like. He worked on arithmetical and multiplicative functions by probabilistic means — probabilistic number theory, the field Rényi and Turán had helped create. And he developed graph-sieve inequalities that operate simultaneously in probability and number theory, along with a long programme on Bonferroni-type inequalities, which bound the probability of a union of events when you know only partial information about their overlaps.
That programme produced *Bonferroni-Type Inequalities with Applications* (1996) with Italo Simonelli, and later *Products of Random Variables* (2004) with the same collaborator. Elsewhere he wrote *Characterizations of Probability Distributions* (1978) with Samuel Kotz, *Introductory Probability Theory* (1984) and *Advanced Probability Theory* (1995). The total came to more than 130 papers and eight books.
Why Janos Is Called a Genius
The specific quality worth naming is unification — the ability to see that apparently unrelated problems share a skeleton. Galambos's Bonferroni inequalities, his graph-sieve results, his work on the digits of real numbers and his extreme value theory all rest on the same underlying instinct: that questions about rare configurations and questions about the structure of numbers are the same questions in different clothing. That instinct is teachable only up to a point, and he clearly had it early; a doctorate under Rényi at twenty-three is not nothing.
The recognition supports a strong reputation without supporting the word. He was elected an external member of the Hungarian Academy of Sciences in 1993 — his homeland claiming him back after two decades abroad — a corresponding member of the Royal Academy of Engineering of Spain in 2001, a member of the International Statistical Institute, and a Fellow of the Institute of Mathematical Statistics.
But the counter-case is clear-eyed. There is no Galambos theorem in the canon. His great book systematised a theory largely built by Fisher, Tippett, Gnedenko and others rather than founding it; he was the field's finest organiser and expositor, which is a different and lesser claim than being its originator. His election to academies came in his fifties and sixties, honouring accumulated contribution rather than a moment of breakthrough. And the tribute his colleagues actually reached for, when he died, was not about brilliance at all: he is remembered as "an inspiring teacher and mentor." That is what the people who knew him chose to put on the record, and it should be allowed to stand as written.
Legacy
Galambos died on 19 December 2019, aged seventy-nine, survived by his wife Éva.
His subject has only grown more urgent. Extreme value theory is now the mathematical foundation of flood defence design, catastrophe reinsurance pricing, structural safety codes and financial risk modelling — every field that has to make decisions about events that have not happened yet and may happen only once. Climate change has made the question of how to reason about unprecedented maxima into one of the defining technical problems of the century.
The man from Zirc who taught in Legon and Ibadan before settling in Philadelphia wrote the book on it. Most of the people using it will never learn his name, which is roughly what happens to mathematics that works.
Achievements
- Held posts at Temple University
- Fields: probability theory, mathematics and mathematical statistics
.jpg)
