Sea stack plots: Replacing bar charts with histograms

Abstract

Graphs in research articles can increase the comprehension of statistical data but may mislead readers if poorly designed. We propose a new plot type, the sea stack plot, which combines vertical histograms and summary statistics to represent large univariate datasets accurately, usefully, and efficiently. We compare five commonly used plot types (dot and whisker plots, boxplots, density plots, univariate scatter plots, and dot plots) to assess their relative strengths and weaknesses when representing distributions of data commonly observed in biological studies. We find the assessed plot types are either difficult to read at large sample sizes or have the potential to misrepresent certain distributions of data, showing the need for an improved method of data visualisation. We present an analysis of the plot types used in four ecology and conservation journals covering multiple areas of these research fields, finding widespread use of uninformative bar charts and dot and whisker plots (60% of all panels showing univariate data from multiple groups for the purpose of comparison). Some articles presented more informative figures by combining plot types (16% of panels), generally boxplots and a second layer such as a flat density plot, to better display the data. This shows an appetite for more effective plot types within conservation and ecology, which may further increase if accurate and user‐friendly plot types were made available. Finally, we describe sea stack plots and explain how they overcome the weaknesses associated with other alternatives to uninformative plots when used for large and/or unevenly distributed data. We provide a tool to create sea stack plots with our R package ‘seastackplot’, available through GitHub.

Potential effects of traffic noise on anuran call characteristics in Louisiana, USA during winter

“…We conducted our analyses in Program R (V. 4.3.1; R Development Core Team, 2023 ) with packages seastackplot (V. 0.0.1.0; Stuart et al., 2024 ), Seewave (V. 2.1.6; Sueur et al., 2008 ), dplyr (V. 1.1.4; Wickham et al., 2023 ), and ggplot2 (V. 3.4.2; Wickham, 2016 ), and exported plots with package ggpubr (V. 0.6.0; Kassambara, 2023 ). We summarized the mean, standard deviation, and range of noise levels for our entire study period and visualized these data using sea stack plots (Stuart et al., 2024 ). We recorded 948 h of acoustic data during our sampling period and randomly selected 1 min per hour to search for anuran vocalizations within our recordings.…”