Designing better figures for small studies: Why you shouldn’t use bar graphs for continuous data and what to do instead (A Visual Q&A Thread)
journals.plos.org/plosbiology/ar…
journals.plos.org/plosbiology/ar…
“Can I still use a bar graph if my data are normally distributed?”
statistika.mfub.bg.ac.rs/interactive-do…
statistika.mfub.bg.ac.rs/interactive-do…
“What should I use instead of a bar graph?”
“Do I need expensive software to do all this?”
No - there are lots of free, user-friendly tools available
No - there are lots of free, user-friendly tools available
“I still prefer bar graphs because they convey a clear message. Sometimes it’s hard to see what’s going on with dot plots.”
Emphasize the data points and de-emphasize the dots to convey a clear message while allowing readers to critically evaluate the data
Emphasize the data points and de-emphasize the dots to convey a clear message while allowing readers to critically evaluate the data
“Can I just add dots to my bar graph?”
This is better than a bar graph, but it won’t be as clear as a dot plot.
This is better than a bar graph, but it won’t be as clear as a dot plot.
“When is it appropriate to show summary statistics like error bars, box plot and violin plots?”
Summary statistics are only meaningful when there is enough data to summarize
Summary statistics are only meaningful when there is enough data to summarize
“Why shouldn’t I use box plots for bimodal data?”
Box plots conceal the two peaks. Bimodal distributions are easier to see with dot plots (smaller samples) or violin plots (large samples)
Box plots conceal the two peaks. Bimodal distributions are easier to see with dot plots (smaller samples) or violin plots (large samples)
“How do I make effective dot plots with box plots or violin plots?”
“Does it matter how I did my stats?”
Yes – The structure of the figure gives the reader information about your experimental design and analysis. Avoid sending mixed messages, especially with small datasets.
Yes – The structure of the figure gives the reader information about your experimental design and analysis. Avoid sending mixed messages, especially with small datasets.
“How do I design figures that match my study design & analysis?”
For simple analyses of small datasets, it’s often clearest to show one graph per analysis. The graph should include all groups, time points, or conditions from the analysis.
For simple analyses of small datasets, it’s often clearest to show one graph per analysis. The graph should include all groups, time points, or conditions from the analysis.
“What can I do to encourage others to use more informative figures in papers, talks and posters?”
“What can I do to increase the likelihood that authors will improve their graphics when I request more informative figures as a reviewer or editor?”
Provide citations to references and resources that will help your colleagues to understand the problem and make better graphics
Provide citations to references and resources that will help your colleagues to understand the problem and make better graphics
