## Ensemble Graphics

Data exploration usually produces many graphics.

Being able to reconstruct these form a command history or notebook is useful.

But a final report or presentation usually needs a small set of carefully chosen graphics. These can be some or all of:

• Several plots of the same type of the same variables; e.g. histograms with different bin widths.

• Differnent views of the same variables, e.g. density plots and box plots.

• Plots of the same type of the same variables for different subgroups; small multiples, or trellis displays.

Many other variations are possible. Such collections are sometimes called ensemble graphics.

• To aid in comparisons, axis ranges encoding choices (color, line type, etc.) should be coordinated.

• Annotations can be used to emphasize important features.

• The graphs should be organized so the most important features can be perceived pre-attentively.

Collections of graphics can be assembled and arranged within the graphics system or using the facilities of the report generation system.

• Within the R graphics system you can use features of par for base graphics and tools from gridExtra for grid-based graphics.

• rmarkdown and LaTeX also provide ways of arranging graphics.

• Tools for creating poster presentations can also help with arranging collections of graphs.

Unwin’s Fig 12.1 provides an ensemble graphic for a data set on the chemical composition of coffee samples:

library(ggplot2)
library(GGally)
## Registered S3 method overwritten by 'GGally':
##   method from
##   +.gg   ggplot2

data(coffee, package = "pgmm")
coffee <- within(coffee, Type <- ifelse(Variety == 1,
"Arabica", "Robusta"))
names(coffee) <- abbreviate(names(coffee), 8)
a <- ggplot(coffee, aes(x = Type)) + geom_bar(aes(fill = Type)) +
scale_fill_manual(values = c("grey70", "red")) +
guides(fill = FALSE) + ylab("")
## Warning: guides(<scale> = FALSE) is deprecated. Please use guides(<scale> =
b <- ggplot(coffee, aes(x = Fat, y = Caffine, colour = Type)) +
geom_point(size = 3) +
scale_colour_manual(values = c("grey70", "red"))
c <- ggparcoord(coffee[order(coffee\$Type), ], columns = 3 : 14,
groupColumn = "Type", scale = "uniminmax") +
xlab("") + ylab("") +
theme(legend.position = "none") +
scale_colour_manual(values = c("grey", "red")) +
theme(axis.ticks.y = element_blank(),
axis.text.y = element_blank())
grid.arrange(arrangeGrob(a, b, ncol = 2, widths = c(1, 2)),
c, nrow = 2)

• consistent coloring across the plots;

• removing some redundant labeling

• having the red lines appear on top in the parallel coordinates plot.

## Information Dashboards

Dashboards are popular in business and arose from work on decision support systems.

Data visualizations are typically a large component of dashboards.

• Some dashboards take the metaphor too far by showing dials that take up lots of space and distract.

• 3D pie charts and the like are also quite common.

• But many are quite effective, convey a lot of information, and emphasize the key items well.

• An example from a paper by Stephen Few:

Dashboards often have dynamic or interactive features:

• the data displayed may be updated on a regular basis (e.g. stock trade activity, current weather conditions);

• the user may be able to interactively change aspects of the visualizations.

Ben Schneiderman’s design guidelines:

• overview first;
• then zoom/filter;
• details on demand.

Jenifer Tidwell’s classification of useful interactions

• scroll and pan;
• zoom;
• open and close;
• sort and rearrange;
• search and filter.

The computational support needed for data updating and interaction will vary.

• Some approaches can be handled by a browser’s JavaScript engine.

• Other approaches need to communicate with a data base or a server process.

## Examples

Tableau provides tools for easily creating dashboards; an example on lyme’s disease in Minnesota.

Rstudio provides the flexdashboard framework for creating dashboards with rmarkdown.

## Some Notes

• Dashboards are popular but limiting.

• More extensive articles, like one on power plants in the US are often more effective.

• Interactive visualizations are also very popular and effective, but often interactive features are not used. Some links to discussions on this can be found here