Home > Math Lint Archive Detail

<< Prev 4/28/2013 Next >>

Statistical Trivia

The prestigious New England Journal of Medicine tried to determine the levels of knowledge of statistics needed to understand its research articles.

NEJM concluded that a reader:

  • Conversant with descriptive statistics could understand 58% of the articles
  • Experience with t-tests increased understanding to 67%
  • Experience with contingency tables raised understanding to 73%
  • Other knowledge (e.g. non-parametric tests, correlation, sample regression, ANOVA, data transformations, multiple regression) raised understanding to 86%
  • Three thoughts. First, what do the percentages even mean? That is, do 58% of the articles contain only descriptive statistics? That is far different from concluding that someone, who has a knowlege of descriptive statistics, is able to translate that into an understanding of an article's use of descriptive statistics.

    Second, does it mean that a reader with knowledge of descriptive statistics will understand 58% of the individual articles (assuming about 15+ articles per issue)...or that they understand 58% of the cummulative content across all of the articles?

    And third, what additional statistics knowledge is needed for the missing 14% of understanding relative to the articles? Or, does this imply that 14% of the articles are not understood by anyone?

    Source: Adapted from Emerson et al, NEJM, 309:709-713