Implement violin charts

[jrmoserbaltimore]

Violin charts are the only type of chart capable of displaying complete distribution data over several cross-sectional data points in a time series, e.g. for displaying the distribution of wages, mean wage, and first/third quartiles per each year.

Useful approach in terms of pgfplotstables:

• A primary data file references other data files
• The primary data file columns are:
• data: the file (relative path ./foo.dat accepted) containing all the plot data
• x: the x axis label
• mean: The mean, calculated if not given or {}
• q1: First quartile, as mean.
• median: Median, as mean.
• q3: Third quartile, as mean.
• std: Standard deviation.
• The data field references a file containing points
• if one column value, all single points
• if two columns count and value, each point is duplicated count times

Note that mean, median, q1, and q3 are given because while pgfplots may be able to compute small data sets, if you're using millions of data points you probably want to compute a few hundred or so representative data points and the mean, median, first and third quartiles, and standard deviation for plotting a decent-resolution chart.

Plots usually represent mean and standard deviation; median and inter-quartile range should also be an option. That part is just a type of box-and-whiskers plotting.

[michaeldorner]

Violin plots in pgfplots would be wonderful. Why? Because they are so more expressive than boxplots:

(taken from https://www.autodesk.com/research/publications/same-stats-different-graphs)

[tapyu]

@michaeldorner How did you make this plot? I need to make violin plots on pgfplots.

[michaeldorner]

I attached the source. But it's not PGF since there is no support yet for them.

[tapyu]

@michaeldorner Yes, I had seen that you wrote in Python... I am afraid this approach is poorly integrated with LaTeX :disappointed:

[tapyu]

@michaeldorner Now there is a solution! Check out my answer!

[michaeldorner]

What an amazing work! Thanks for sharing!!! ❤️