Statistical graphs are commonly used in scientific publications. Unfortunately, graphs in psychology journals rarely portray distributional information beyond central tendency, and few graphs portray inferential statistics. Moreover, those that do portray inferential information generally do not portray it in a way that is useful for interpreting the data. The authors present several recommendations for improving graphs including the following: (a) bar charts of means with or without standard errors should be supplanted by graphs containing distributional information, (b) good design should be used to allow more information to be included in a graph without obscuring trends in the data, and (c) figures should include both graphic images and inferential statistics presented in words and numbers.
Lane et al. (Thu,) studied this question.