ABSTRACT
It has been more than 15 years since Efron’s (1979) monumental paper on bootstrap methods. This paper has had a major impact on the field of statistics. In Kotz and Johnson’s (1992) book Breakthroughs in Statistics, which selected, highlighted, and reprinted the most significant works in statistics since the 1980s, Efron’s (1979) paper is the latest one selected. Therefore, although Efron’s paper is quite “young” relative to the history of statistics, its significance has already been well recognized. The beauty of the bootstrap, of course, lies in its relaxation of severe distributional assumptions of parametric models (e.g., multivariate normality of the popu lation distribution). Statistical inferences using the bootstrap are usually made by computations, thus providing solutions to statistical problems that would othervvise be intractable. The bootstrap has been applied to many areas in statistics, and a vast amount of research on it has been published since Efron’s introduction. The bootstrap also has diffused into the field of behavioral sciences, though at a much slower pace. In the psychological literature, there has been a heated debate about the usefulness of the boot strap applied to the correlation coefficient (cf. Efron, 1988; Lunneborg, 1985; Rasmussen, 1987, 1988; Strube, 1988). A recent application has been tied to coefficients in discriminant analysis (Dalgleish, 1994). In the socio logical literature, to the best of our knowledge, the earliest introduction of the bootstrap methodology into the field can be traced back to the work byBollen and Stine (1988), Dietz, Frey, andKalof (1987), and Stine (1989). Although not primarily written for social scientists, articles and books also
exist that introduce the basic bootstrap ideas in a way comprehensible to social scientists with minimal formal statistical training (e.g., Efron Se Gong, 1983; Efron Se Tibshirani, 1986, 1993; for a recent critical review of the bootstrap research, see Young, 1994).
