Suppose, for example, we were interested in how PTSD symptoms differed between male and female combat veterans. In addition to enabling the rapid understanding of information, visuals also provide a critical quality-control check. All of these reasons lead to massive incentives to utilize misleading, suboptimal visualizations or produce no graphics at all. In nearly all cases, creating visuals is ancillary to the analysis individuals must perform additional steps, often clicking through deeply nested menus in order to produce a graphic (Hu, Orghian, & Hidalgo, 2018). Still others (e.g., R) have such a steep learning curve that only the most savvy of researchers even attempt to pilot its code. Other programs (e.g., SPSS) make it difficult or impossible to produce graphs that follow the practices we outline for data visualization, such as plots with raw data when group means are plotted. Furthermore, some statistical programs (e.g., Excel), make it extremely difficult or even impossible to produce standard graphics (e.g., violin plots, boxplots, or even histograms). Also, because the bottoms of bar graphs fall at y = 0, they give the false impression negative values are impossible, which may lead some to underestimate the variability in data. Representing means using bar graphs leads people to believe values within the bar are more likely than values above the bar, even for symmetric distributions (Zacks & Tversky, 1999). For example, past surveys of psychological research show that both textbooks (Butler, 1993 Peden & Hausmann, 2000) and journal articles (Kyonka et al., 2019 Schild & Voracek, 2013 Weissgerber, Milic, Winham, & Garovic, 2015) rely heavily on graphics that bias human perception, e.g., bar graphs and line graphs. Footnote 1 While there are some programs with light coding requirements and sensible defaults (D3, VegaLite, Spotfire, Tableau, Power BI, etc.), these programs and the visualization principles they follow are not well adopted in psychological (or biomedical) research. In case you have any further questions, let me know in the comments.We agree, at least with the most popular statistical software platforms used in psychology. However, we could easily extend the previous R codes to plot more complex graphics such as grouped or stacked barcharts. Note that we have used a relatively simple example in this tutorial.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |