The New Statistics for applied linguistics

Abstract

The New Statistics is an approach to scholarly research which offers an alternative to the problematic overreliance on significance testing currently plaguing the research literature. This paper describes the problems associated with significance testing and introduces the key concepts of the data-analysis that best fits with the goals of the New Statistics: estimation of effect sizes and confidence intervals. These concepts will be applied in a reanalysis of the summary data from an article that was recently published in this journal. This makes it possible to compare the estimation approach advocated by the New Statistics to the standard significance tests and to discuss potential drawbacks of this approach as a means of gathering quantitative evidence in support of our substantive hypotheses.