In this post I have reviewed some of the most common interactive visualization packages in R with simple example plots along with some comments and experiences. Here are the packages included:
- ggplot2 – one of the best static visualization packages in R
- ggvis – interactive plots from the makers of ggplot2
- plotly – convert ggplot2 figures to interactive plots easily
- googleVis – use Google Chart Tools from R
You can either jump straight to the example visualization or read my comments first. The R markdown source code for this blog post with embedded visualizations can be found in Github. I have probably missed some important features and documentation, and also clear mistakes are possible. Please point those out in the comments, and I’ll fix them. It is also important to note that I am a heavy ggplot2 user, and hence my comments may also be biased!
Also other libraries for creating interactive visualizations from R do exist, such as clickme, RIGHT, ggobi, iplots, gg2v, rVega, cranvas and r2d3. Some of these are not under active development anymore. I might include some of those into the comparison here in the future. The d3Network package is also worth checking if you need cool interactive network visualizations.
%>% (familiar for dplyr users), replacing the
All other packages than googleVis are clearly in an early development phase, which is visible in a limited features and documentation. As an experienced ggplot2 user it was often hard to adapt to the much narrower range of features included in ggvis. For example faceting is a very important feature that hopefully gets implemented soon.
Documentation-wise ggvis and googleVis seem to be the most advanced. rCharts especially suffers from the combination of multiple plot types (named rather uninformatively as
nPlot and so on) with practically no documentation. So producing anything else than what’s provided in the existing examples was very hard.
googleVis sets itself apart by requiring the data in a different format than the other packages. In Hadley Wickham’s terms, it assumes the data is in the messy format, in contrast to the other packages, tha assume tidy data. This makes it somewhat hard to use, at least when one is used to using tidy data frames. See the examples below for more details.
Plotly is an interesting alternative to the other packages in that it simply takes as input a ggplot2 object and transforms it into an interactive chart that can then be embedded into websites. Using the service requires authentication, which is a clear limitation. By default all plots are made publicly visible to anyone, but there apparently is a way to produce private plots as well, with a limit in their number in the free account.
ggvis is currently the only one of these packages that can not produce map visualizations, but I assume this feature will be added in the future. plotly can use maps created with ggplot2, but not yet with the handy ggmap extension.
Sharing the visualizations
Interactive visualizations are typically meant to be shared for a larger audience. Common ways to share interactive visualizations from R are as standalone html files, embedded in R markdown documents, and embedded in Shiny applications. All the studied packages can produce standalone htmls, though possibly with some loss of interactivity.
R markdown documents are a very nice way of sharing reproducible analyses, using the knitr and rmarkdown packages. Outputs from all the studied visualization packages can be embedded in .Rmd documents, though I had some problems (see the Issues section below). All packages are also compatible with Shiny documents and applications, and have good tutorials for this.
I encountered several problems when preparing this blog post. Specifically, I had issues in embedding the plots into R markdown documents. This is made more complicated with the various available ways of turning .Rmd files into html: manual
knit2html() functions, the
Knit HTML button in RStudio, and a Jekyll-powered blog with its own requirements. Here I have listed the most important issues, with solutions when found. Some things are still unsolved, hope someone can help me with those!
- ggvis showed up nicely with
- rCharts: Embedding rCharts to R markdown did not quite work either as shown e.g. here. With
Knit HTMLbutton the line that worked was
rchars.object$print(include_assets=TRUE), whereas with the blog the line was
- plotly: Embedding plotly charts into R markdown documents did not work as shown here, but adding
ggplotly()call solved the issue (thanks to Scott Chamberlain and Marianne Corvellec for help!). Note that in this post I embedded existing plotly charts manually.
I also noticed some minor issues:
- googleVis was missing axis labels by default
- rCharts is missing legend titles, and behaves strangely on scatter plot: legend shows partially incorrect information, and the plot area is too tight
## Install necessary packages install.packages("devtools") library("devtools") install.packages("ggvis") install.packages("googleVis") install_github("ramnathv/rCharts") install_github("ropensci/plotly") install.packages("dplyr") install.packages("tidyr") install.packages("knitr")
# Load packages library("ggvis") library("googleVis") library("rCharts") library("plotly") library("dplyr") library("tidyr") library("knitr") # Define image sizes img.width <- 450 img.height <- 300 options(RCHART_HEIGHT = img.height, RCHART_WIDTH = img.width) opts_chunk$set(fig.width=6, fig.height=4)
Plotly needs some setting up (using the credentials from here).
# Plotly requires authentication py <- plotly("RgraphingAPI", "ektgzomjbx")
Prepare the mtcars data set a bit.
# Use mtcars data data(mtcars) mtcars$cyl <- factor(mtcars$cyl) mtcars$am <- factor(mtcars$am) # Compute mean mpg per cyl and am mtcars.mean <- mtcars %>% group_by(cyl, am) %>% summarise(mpg_mean=mean(mpg)) %>% select(cyl, am, mpg_mean) %>% ungroup()
hist.ggplot <- ggplot(mtcars, aes(x=mpg)) + geom_histogram(binwidth=1) hist.ggplot
hist.ggvis <- mtcars %>% ggvis(x = ~mpg) %>% layer_histograms(width=1) %>% set_options(width = img.width, height = img.height) hist.ggvis
# rCharts histogram needs manual binning and counting! hist.rcharts <- rPlot(x="bin(mpg,1)", y="count(id)", data=mtcars, type="bar") # Use this with 'Knit HTML' button # hist.rcharts$print(include_assets=TRUE) # Use this with jekyll blog hist.rcharts$show('iframesrc', cdn=TRUE)