“Sleeping Beauties” in Statistics
Nozer D. Singpurwalla · Significance · 2020
In 2015, a paper was published that ranked the top 15 “sleeping beauties in science”.1 The term “sleeping beauty” was first adopted by Ton van Raan in 2004 to describe a publication “that goes unnoticed … for a long time and then, almost suddenly, attracts a lot of attention”,2 meaning citations. Building on the work of van Raan and others, the authors of the 2015 article proposed a parameter-free statistical approach to quantify the extent to which a paper can be considered a sleeping beauty, using an index, B – a “beauty coefficient”. Through the use of B, the authors identified the field of statistics as a top producer of sleeping beauties. Admittedly, of the 15 papers discussed in the article, only two are statistics papers: Karl Pearson's 1901 paper, “On Lines and Planes of Closest Fit to System of Points in Space”,3 ranked sixth; and E. B. Wilson's 1927 paper, “Probable Inference, the Law of Succession, and Statistical Inference”,4 ranked 10th. However, in supplementary material, the authors identified a further 10 statistics papers with high beauty coefficients, including those pertaining to Fisher's exact test, the Metropolis–Hastings algorithm, and Kendall's rank correlation coefficient (bit.ly/SBsupp). The 2015 article also provided an approach for identifying the “awakening time” of a sleeping beauty – that is, the year when citations begin to explode – and the authors concluded that many sleeping beauties wait more than 50 years. This was certainly the case for Pearson's article, which awakened in 2002 after a century of sleep, while the Wilson paper waited 72 years, until 1999, for its wake-up call. Why the delay? The Pearson paper, now almost 120 years old, pertains to the thesis that when considering a relationship between two variables x and y, with both x and y observed with error, the regression of x on y need not be the same as the regression of y on x, and that the line of best fit, as measured in terms of minimum sum of squares, is neither of these regressions but the sum of squares of the perpendicular distances between the points and the line. The message of the paper is relatively easy to appreciate. But ploughing through its details is something of chore. Perhaps this demand on time and patience, plus the fact that its publication was in an unconventional (philosophical) outlet, could be the cause of its deep slumber. And what of the Wilson paper? It pertains to the matter of appreciating the meaning of a probability interval, something articulated by Wilson5 before Neyman's introduction of a confidence interval in 1937. Wilson's style of exposition could be a possible cause of his paper's delayed recognition – he is very descriptive and verbose – while the paper's sudden awakening could be due to an explosion of interest in the score interval for biostatistics applications. Alas, it can be difficult to pin down exactly why a paper becomes a sleeping beauty. But, given a slumber time of 50 years or more, it seems likely that authors of many such papers will be in a permanent state of sleep themselves when their beauties finally wake. Rock'n Roll Monkey/Unsplash.com “My name is Artemis, and I am a bot. I monitor the time series of hundreds of stocks, looking for known patterns in share price changes in order to make better investment decisions. I would like to tell you how I do that.” Artemis's story, as “told” to Marco Antonio Andrade Barrera, was a runner-up in our 2019 writing competition. significancemagazine.com/641 A headline in The Times in February claimed that a “[r]ound of golf once a month can reduce risk of early death”. But, as Full Fact's Abbas Panjwani points out, the study this claim is based on did not control for economic status, so “the research could simply be affirming something we already know – that being rich is good if you want to live longer”. bit.ly/3c3TNhE CDC/Unsplash.com From the Significance archive: Media reports of the dangers of swine flu can leave much to be desired. But then scientific reports can be just as unclear. Diamanto Mamuneas and Maria Viskaduraki have sympathy for two journalists and help a frightened friend. bit.ly/2SEQwNh The full Significance archive is at bit.ly/SIGNarchive. Magazine articles older than one year are free to read. Articles less than a year old require a subscription, or ASA or RSS membership. See signiflcancemagazine.com/subscribe for more information. Visit significancemagazine.com for exclusive articles, magazine extras and links to the online version of this issue.