How to create a violin plot with ggplot2 in R with geom_violin and geom_dotplot (CC092)

One of the trendy data visualization methods is a violin plot. In this Code Club, Pat will show you how to create a violin plot in ggplot2 with geom_violin and how to show individual data points on the plot with ggplot2's geom_jitter and geom_dotplot functions. We'll also see how to indicate specific quantiles on the plot.

Do you have a figure that you would like to receive a critique or help improving? Let me know and I'd be happy to arrange a guest appearance!

You can also find complete tutorials for learning R with the tidyverse using...

0:00 Introduction
2:01 Creating a violin plot
5:16 Adding individual data to violins w/ geom_jitter
6:02 Adding individual data to violins w/ geom_dotplot
9:23 Adding quantiles to violin plot
11:01 Adding median with stat_summary
11:53 Removing violin body
12:59 Critique of violin and box plots
17:09 Conclusion

[bridgettsmith]

Thank you for this thorough explanation

[ianworthington]

A very interesting video. I've been favouring violins (actually half violins) over box plots for a while now, for the same reason that boxes are better than just the mean and sd. But I've always been a little nervous about the impact of the mysterious "kernel smoothing" algorithm, and this video seems to show that clearly. I might have to investigate the dot plot histogram instead. Many thanks.

[jessicabrassard]

I am working on a final project for a stats class I am taking at Michigan Tech, and clicked on this video for pointers on creating my my violin plots. The first thing I noticed is your hat! Go (tokyodachi) huskies!

[cyg]

I find combining a boxplot within a violin plot is a good solution, "ggridges" can be good to compare distributions too.

[nicolasberiot]

Thanks for the video! For me violin are great when you want to say something about different distributions but in most cases you are interesting in comparing the mean or median so boxplots would be better.

[carllessard]

Great video and critics on the violin plot. What about merging the boxplot with the violin? It may help a reader to improve the visualization of the distribution of the data. Just my two cents comment

[oluwaseyeoladele]

Thanks for this, found it really helpful. Quick question please, is it possible to plot a violin plot with individual data points and also add the mean as a crossbar or in other geom forms? thanks ahead

[aleksandrarzamasov]

Thank you for the video! I don't always get why violin plots are considered visually intuitive. They do look great when data is normally distributed. However, it becomes a bit hard to compare violins plots for distributions like in the video example.

Btw, I am curious, is there an "optimal" way to visualize count matrices to emphasize clustering (in my case, binary matrices and matrices with small count numbers, e.g., distribution of specific genes across bacterial genomes)? I've seen people mostly doing heatmaps and sometimes PCA plots, but still curious what way is considered the best.

[xxXXCarbonXXxx]

Box plot is the clear win, however sometimes entertainment trumps information. I'd never use one in a paper but maybe for a team talk I would use one (assuming a box plot wasn't available)