I&#39;d like to create a split violin density plot using ggplot, like the fourth example on [this page](https://seaborn.pydata.org/generated/seaborn.violinplot.html) of the seaborn documentation.

[![enter image description here][1]][1]

Here is some data:

    set.seed(20160229)

    my_data = data.frame(
        y=c(rnorm(1000), rnorm(1000, 0.5), rnorm(1000, 1), rnorm(1000, 1.5)),
        x=c(rep(&#39;a&#39;, 2000), rep(&#39;b&#39;, 2000)),
        m=c(rep(&#39;i&#39;, 1000), rep(&#39;j&#39;, 2000), rep(&#39;i&#39;, 1000))
    )


I can plot dodged violins like this:

    library(&#39;ggplot2&#39;)

    ggplot(my_data, aes(x, y, fill=m)) +
      geom_violin()

[![enter image description here][2]][2]

But it&#39;s hard to visually compare the widths at different points in the side-by-side distributions. I haven&#39;t been able to find any examples of split violins in ggplot - is it possible?

I found a [base R graphics solution](https://gist.github.com/mbjoseph/5852613) but the function is quite long and I want to highlight distribution modes, which are easy to add as additional layers in ggplot but will be harder to do if I need to figure out how to edit that function.


  [1]: https://i.stack.imgur.com/tYQWX.png
  [2]: http://i.stack.imgur.com/8De26.png

All Violin Plot Solutions on Gang of Coders

Total of 1 Violin Plot Solutions

Split violin plot with ggplot2