Jump to content

Talk:Phillip Good

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Untitled

[edit]

My name is Michael Chernick. I knew Phil Good when I lived in Southern California in the 1980s and 1990s. We first met when I invited him to give a talk on resampling methods at a local chapter meeting of SCASA the Southern California Chapter of the American Statistical Association. Some of the biographical information in this biography is accurate but I object to his claim that his applied paper in 1975 was the first application of the bootstrap. At that time Brad Efron had already started research on the bootstrap and was working on his famous paper that was first published as a Stanford University technical report in 1977 and after the refereeing process his seminal paper in the Annals of Statistics where he first named the method bootstrap. Phil once related to me the story about how he discussed his paper with Efron at a meeting sometime in 1975 and after hearing the description Efron told him that it sounded much like his new method. In reality claims of priority are shaky at best. Julian Simon on his own and later in collaboration with Peter Bruce was using the bootstrap approach as a teaching tool in the 1960s. Efron admits that his idea was an exrension of work done by Hartigan in the early 1970s on subsampling. I discuss these historical events in the first chapter of my book Bootstrap Methods published by Wiley in 1999 and in a second edition in 2007. The jackknife is a cousin to the bootstrap as also are permutation methods. Permutation methods in statistics go back at least to Fisher and Pitman in the late 1930s. The jackknife came about in the late 1940s in the work of Quenouille. So these close relatives of the bootstrap were know well before 1975 and certainly Simon has documentation of his use of the Monte Carlo approximation to the bootstrap method in his 1969 book. The Monte Carlo approximation is such a natural technique that it is very likely that many people have independently thought of it and tried it but never puvlished anything about it. So I view the claims that Simon and now Good make about being the first are suspect and unimportant. Efron deserves credit for originating the bootstrap because he constructed the beginnings of a theory for it and showed its close relationship to the jackknife. Prior to 1979 the viability of the bootstrap as a technique with good statistical properties was unknown. Even after 1979 its true value was not appreciated until more of the theory and successful applications were established. The flood of literature by Efron, Tibshirani, Beran, Bickel, Freedman and others can only be attributed to Efron's work, The work of Simon and Good did not spawn any such movement in the statistical community.