Doug Altman – A Giant Among Statisticians

Photo by NDORMS, University of Oxford

I’m sure there will be many obituaries over the coming days, and if you want a synopsis of Professor Altman’s accomplishments you can read his Wikipedia entry. Here, I provide a few personal reflections on a great man.

Thirty years ago, the Imperial Cancer Research Fund[i] was a relatively small research organisation employing some of the world’s leading cancer researchers. Somewhat unusually its faculty included (at least) four statisticians who each have made a huge contribution to medical research. The Director General was the geneticist, Sir Walter Bodmer. Walter’s first degree was in mathematics; he completed his doctorate with Sir Ronald Fisher (“the father of modern statistics”); and in 1984-85 he was president of the Royal Statistical Society. Sir Richard Peto started his career publishing statistical papers but went on to become one of the leading epidemiologists in the world. Jack Cuzick’s early publications were in mathematical statistics and probability theory, but he went on to be one of the pioneers of therapeutic cancer prevention. The newest recruit was Doug Altman, an applied statistician whose impact on medical research has been colossal.

When I started work I was privileged to be in Jack Cuzick’s team, sharing the same office building as Doug Altman. Doug did not limit the scope of applied statistics to “data analysis”. Rather he had a vision for statistics to revolutionise the quality of medical research as a whole – in 1994 he wrote an editorial in the BMJ on “The scandal of poor medical research”. His approach was not only to lead by example but also to teach, write, and produce guidelines or “recipes” for improving the standards of statistics and statistical reporting that have been adopted by all leading medical journals.

Doug tackled the problem of teaching statistics to medical students and researchers head on, first through books – his 1990 “Practical Statistics for Medical Research” is still a classic – and later through his “Statistics Notes” in the BMJ. The success of his book can be understood from his read paper (with Martin Bland) to the Royal Statistical Society in 1991 which I remember to this day. First, he analysed the problem. Quoting from Michael Healy he said “I do not know a single discipline other than statistics in which it is a positive recommendation for a new textbook, worthy of being quoted on the dust cover, that it is not written by a specialist in the appropriate field. Would any medical reader read, would any medical publisher publish, my new introduction to brain surgery – so much simpler and more clearly written than those by professional brain surgeons, with their confusing mass of detail? I trust not.” Then he researched the topic well. He read umpteen statistical textbooks and found glaring errors such as one in which the author applied two equivalent tests that should give identical P-values. The author made a mistake in his calculations, so one was significant but the other was not. Instead of realising that he must have made a mistake he argued that one test was more sensitive than the other because it requires fewer assumptions. With such nonsense in textbooks, was it any wonder that the general standard of statistics in medical research was so poor.  Systematically, Doug spent much of his career changing that through articles, checklists and consensus statements.

A friend studying for membership of the Royal College of Obstetrics and Gynaecology was thrilled to learn that I worked with Doug. “His textbook is like no other I have come across. With a little work I can read and understand it, and every example comes from a published medical article, so they make sense and are interesting.” Some years later a medical colleague considering doing a course in medical statistics asked: Who’s teaching it? She had once been to a lecture by Doug Altman and subsequently was only willing to take the course if he was on the faculty! But it was not just medics who admired Doug’s book. I vividly recall Doug sharing with me a document he had received from two eminent medical statisticians in Copenhagen. They had gone through Practical Statistics for Medical Research with a fine tooth-comb. They had not of course, found any real mistakes, but several small typos and they posed a few challenging questions regarding details of the text and whether Doug’s simple and understandable choice of analysis was optimal for a particular example in the book. Others might have been annoyed to receive such a pedantic review, but Doug was honoured that they had taken the time to read his book so carefully. We worked through the challenging questions discussing whether anything should change for the second edition. I was quite junior, and it was his book, but he treated me as an equal.

Of the four statisticians I mentioned earlier, Doug is the only one not to have be made a Fellow of the Royal Society, yet his work has been cited over 200,000 times –more than the other three great statisticians’ citations combined. In terms of impact on medical research his contribution is immense.

I never published with Doug Altman[ii], but for many years I would chat to him almost daily about something either he or I was working on. Doug’s views were always relevant, and I left his office wiser. I miss those chats, his fatherly advice and sound counsel. But if I close my eyes I’m sure I’ll still be able to hear his advice when I really need it.

You may be interested in the Doug Altman Scholarship

[i] Forerunner of Cancer Research UK

[ii] Not unless you count a note in the Stata Technical Bulletin

The views expressed are those of the author. Posting of the blog does not signify that the Cancer Prevention Group endorse those views or opinions.

Subscribe to our mailing list to get updates of new posts…

Be the first to comment

Leave a Reply

Your email address will not be published.