R: How to overlay pie charts on 'dots' in a scatterplot in R

14
votes

Using R I would like to replace the points in a 2d scatter plot by a pie chart displaying additional values.

The rational behind is that I have time series data for hundreds of elements (proteins) derived from a biological experiment monitored for 4 conditions. I would like to plot the elements (categorial data) on the y axis and occurrence of a event in time on the x axis. To visualize the relative occurrence between the 4 conditions I would like to visualize this in form of a pie chart or doughnut chart overplayed onto the respective point in the scatter plot. The overall data density is low so overlapping won't be an issue.

Is this possible in R? I was thinking of using a manual scale in ggplot2 but could not figure out how to define a pie chart as a scale.

Also of interest would be how to best cluster this data and sort it accordingly.

2
rchartsggplot2visualizationcluster-analysis
I think that to do it in ggplot you would have to write a geom_pie -- at present pie charts in ggplot are implemented via a coordinate transformation, which only works if you want to plot a single pie per facet ... – Ben Bolker
The scatterpie package can do this: cran.r-project.org/web/packages/scatterpie/vignettes/… – Arthur Yip

2 Answers

10
votes

Yes. pieGlyph() is one ready-to-go function from the Rgraphviz package.

Also, I would check out this Q/A for how to do things like this more generally:

How to fill a single 'pch' point on the plot with two-colours?

Especially check out ?my.symbols from the TeachingDemos package.

Lastly, in regards to ggplot2, you should check out this blog post about possible upcoming features:

http://blog.revolutionanalytics.com/2011/10/ggplot2-for-big-data.html

3
votes

See also Paul Murrell. Integrating grid graphics output with base graphics output. R News, 3(2):7-12, October 2003. http://www.r-project.org/doc/Rnews/Rnews_2003-2.pdf

The code on pp 10-11 sets up the main plot axes, labels and legend, and then opens a series of smaller windows centered at each individual point on the plot and plots a small graph in each window. I've tried pie charts, mosaics and barplots, but the method is not limited to these types.