Judging statistical significance by inspection of standard error bars

Abstract

This paper describes the evaluation of statistical significance by visual comparison of the distance between means relative to the lengths of standard error bars in graphic representations of data. The relationship between the t statistic and means separated by various numbers of standard error bars permits the observer to make quick conservative judgments of the significance of differences between means without computations or extensive tables. When standard error bars touch or overlap, the means do not differ significantly. Means that differ by three times the length of a standard error bar differ significantly at the.05 level as long as the sample size per mean is nine or more. © 1989, The Psychonomic Society, Inc.. All rights reserved.