When One Shape Does Not Fit All: A Commentary Essay on the Use of Graphs in Psychological Research

Frontiers in Psychology

Massimiliano Pastore¹ 

关键词: statistical reasoning;    bar chart and box plot;    graphical representation;    exploratory data analysis;    credibility crisis;   

DOI  :  10.3389/fpsyg.2017.01666

学科分类：心理学（综合）

来源: Frontiers

【 摘 要 】

A recent systematic review (Weissgerber et al., 2015) of research articles published in top physiology journals in 2014 showed that the most often used graphical representations are static graphs and, among these, the widely known bar chart. Bar charts, useful for depicting frequencies and the occurrence of categorical variables, summarize means and standard deviations without depicting the underlying distribution of data. It results that nothing else is provided beyond what already the statistics show, increasing the risk to misinterpret research findings and to not detect important information (Cooper et al., 2002; Schriger et al., 2006; Saxon, 2015; Gelman, 2017). For example, the presence of anomalous outliers or of marked asymmetry cannot be inferred. Though a systematic review on this topic has not been published yet in psychology, research articles published between January and June 2016 in four high impact psychology journals (Behaviour Research Methods, Cognitive Psychology, Psychological Science, and Trends in Cognitive Science) suggest a state of the art that does not differ much from that of other disciplines using statistical methods, with a significant presence of bar chart graphs in a field of research where continuous variables (e.g., reaction times, psychological test scores) are almost the norm (see also Bar Bar Plots Project, 2017). Specifically, on 131 research papers examined, bar charts were about 55% of 104 presented graphs.

【 授权许可】

CC BY