当前位置: 首页 > 期刊 > 《新英格兰医药杂志》 > 2005年第18期 > 正文
编号:11328688
Statistical Methods in the Journal
http://www.100md.com 《新英格兰医药杂志》
     To the Editor: Previous surveys, in 1978 to 1979 and in 1989, of Original Articles published in the New England Journal of Medicine revealed increasing sophistication of statistical methods over time.1,2 The use of more sophisticated statistical methods may challenge clinicians' comprehension of new research and could slow dissemination of study results.

    We have updated the findings with data from 311 articles published in volumes 350 through 352 (January 2004 through June 2005). For each article, we reviewed the Methods section and scanned the Results section and supplemental appendixes for statistical content according to the set of categories used in previous reviews that, in many cases, we augmented with new methods. Each article was independently coded by both of us. Any discrepancy between these reviews was resolved with the use of a consensus process.

    Table 1 shows the frequency of statistical methods used in the Original Articles published in the Journal over time. The percentage of articles containing no statistical methods at all or only simple descriptive statistics did not change substantially between the 1989 study and the present study. There were substantial increases between the 1989 study and the present study in the percentages of articles that included contingency tables, epidemiologic statistics, survival methods, multiple regression, multiple comparisons, and power analyses.

    Table 1. Statistical Content of Original Articles in the New England Journal of Medicine over Time.

    The "accumulation by article" column in Table 1 made use of a hierarchical categorization of methods of increasing complexity, as outlined by Emerson and Colditz.1,2 For example, the entry for contingency tables shows that only 47 articles (15 percent) used no methods beyond descriptive statistics, used t-tests, and used contingency tables. Similarly, if a reader had knowledge of t-tests, contingency tables, nonparametric tests, epidemiologic statistics, Pearson's correlation, simple linear regression, analysis of variance, transformations, and nonparametric correlation (topics typically included in introductory statistics courses), then 21 percent of the articles would be accessible, as defined by Emerson and Colditz. The biggest jumps in the "accumulation by article" percentage related to knowledge of multiple regression, power, repeated-measures analysis, and missing-data methods.

    More than half the articles used relatively sophisticated statistical methods such as survival analysis or multiple regression. Consistent with previous research, our findings showed that there was a continued trend toward increased use of newer and more complex methods not typically included in introductory or second-level statistics courses. We concur with Emerson and Colditz's statement from 1992 that "an acquaintance with a few basic statistical techniques cannot give full statistical access to research appearing in the Journal,"2 and we believe that this is even more true in 2005. The increasing sophistication of statistical methods has potential implications for medical and statistical educators.3,4,5

    Nicholas J. Horton, Sc.D.

    Suzanne S. Switzer

    Smith College

    Northampton, MA 01063-0001

    nhorton@email.smith.edu

    References

    Emerson JD, Colditz GA. Use of statistical analysis in the New England Journal of Medicine. N Engl J Med 1983;309:709-713.

    Emerson JD, Colditz GA. Use of statistical analysis in the New England Journal of Medicine. In: Bailar JC III, Mosteller F, eds. Medical uses of statistics. 2nd ed. Waltham, Mass.: NEJM Books, 1992:45-57.

    Moore TL. Teaching statistics: resources for undergraduate instructors. Washington, D.C.: Mathematical Association of America, American Statistical Association, 2000.

    Altman DG. Statistics in medical journals: some recent trends. Stat Med 2000;19:3275-3289.

    Altman DG, Bland JM. Improving doctors' understanding of statistics. J R Stat Soc 1991;154:223-67.