pirateplot

Nathaniel Phillips

2016-10-24

What is a pirateplot()?

A pirateplot, is the RDI (Raw data, Descriptive statistics, and Inferential statistics) plotting choice of R pirates who are displaying the relationship between 1 to 3 categorical independent variables, and one continuous dependent variable.

A pirateplot has 4 main elements

  1. points, symbols representing the raw data (jittered horizontally)
  2. bar, a vertical bar showing central tendencies
  3. bean, a smoothed density (inspired by Kampstra and others (2008)) representing a smoothed density
  4. inf, a rectangle representing an inference interval (e.g.; Bayesian Highest Density Interval or frequentist confidence interval)

Main arguments

Here are the main arguments to pirateplot()

Main Pirateplot Arguments
Argument Description Examples
formula A formula height ~ sex + eyepatch, weight ~ Time
data A dataframe pirates, ChickWeight
main Plot title ‘Pirate heights’, ’Chicken Weights
pal A color palette ‘xmen’, ‘black’
theme A plotting theme 0, 1, 2
inf Type of inference ‘ci’, ‘hdi’, ‘iqr’

Themes

pirateplot() currently supports three themes which change the default look of the plot. To specify a theme, use the theme argument:

Theme 1

theme = 1 is the default

# Theme 1 (the default)
pirateplot(formula = weight ~ Time,
           data = ChickWeight,
           theme = 1,
           main = "theme = 1")

Theme 2

Here is theme = 2

# Theme 2
pirateplot(formula = weight ~ Time,
           data = ChickWeight,
           theme = 2,
           main = "theme = 2")

Theme 3

And now…theme = 3!

# Theme 3
pirateplot(formula = weight ~ Time,
           data = ChickWeight,
           theme = 3,
           main = "theme = 3")

Theme 0

theme = 0 allows you to start a pirateplot from scratch – that is, it turns of all elements. You can then selectively turn elements on with individual arguments (e.g.; bean.f.o, point.o)

# Default theme
pirateplot(formula = weight ~ Time,
           data = ChickWeight,
           theme = 0,
           main = "theme = 0\nStart from scratch")

Color palettes

You can specify a general color palette using the pal argument. You can do this in two ways.

The first way is to specify the name of a color palette in the piratepal() function. Here they are:

piratepal("all")

For example, here is a pirateplot using the "pony" palette

pirateplot(formula = weight ~ Time,
           data = ChickWeight,
           pal = "pony", 
           theme = 1,
           main = "pony color palette")

The second method is to simply enter a vector of one or more colors. Here, I’ll create a black and white pirateplot from theme 2 by specifying pal = 'black'

pirateplot(formula = weight ~ Time,
           data = ChickWeight,
           theme = 2,
           pal = "black",
           main = "pal = 'black")

Customising elements

Regardless of the theme you use, you can always customize the color and opacity of graphical elements. To do this, specify one of the following arguments. Note: Arguments with .f. correspond to the filling of an element, while .b. correspond to the border of an element:

Customising plotting elements
element color opacity
points point.col, point.bg point.o
beans bean.f.col, bean.b.col bean.f.o, bean.b.o
bar bar.f.col, bar.b.col bar.f.o, bar.b.o
inf inf.f.col, inf.b.col inf.f.o, inf.b.o
avg.line avg.line.col avg.line.o

For example, I could create the following pirateplots using theme = 0 and specifying elements explicitly:

pirateplot(formula = weight ~ Time,
           data = ChickWeight,
           theme = 0,
           main = "Fully customized pirateplot",
           pal = "southpark", # southpark color palette
           bean.f.o = .6, # Bean fill
           point.o = .3, # Points
           inf.f.o = .7, # Inference fill
           inf.b.o = .8, # Inference border
           avg.line.o = 1, # Average line
           bar.f.o = .5, # Bar
           inf.f.col = "white", # Inf fill col
           inf.b.col = "black", # Inf border col
           avg.line.col = "black", # avg line col
           bar.f.col = gray(.8), # bar filling color
           point.pch = 21,
           point.bg = "white",
           point.col = "black",
           point.cex = .7)

If you don’t want to start from scratch, you can also start with a theme, and then make selective adjustments:

pirateplot(formula = weight ~ Time,
           data = ChickWeight,
           main = "Adjusting an existing theme",
           theme = 2,  # Start with theme 2
           inf.f.o = 0, # Turn off inf fill
           inf.b.o = 0, # Turn off inf border
           point.o = .2,   # Turn up points
           bar.f.o = .5, # Turn up bars
           bean.f.o = .4, # Light bean filling
           bean.b.o = .2, # Light bean border
           avg.line.o = 0, # Turn off average line
           point.col = "black" # Black points
           )

Just to drive the point home, as a barplot is a special case of a pirateplot, you can even reduce a pirateplot into a horrible barplot:

pirateplot(formula = weight ~ Time,
           data = ChickWeight,
           main = "Reducing a pirateplot to a barplot",
           theme = 0, # Start from scratch
           bar.f.o = .7) # Just turn on the bars

Additional arguments

There are several more arguments that you can use to customize your plot:

Additonal pirateplot elements
element arguments examples
Background color back.col back.col = ‘gray(.9, .9)’
Gridlines gl.col, gl.lwd, gl.lty gl.col = ‘gray’, gl.lwd = c(.75, 0), gl.lty = 1
Quantiles quant, quant.lwd, quant.col quant = c(.1, .9), quant.lwd = 1, quant.col = ‘black’
Average line avg.line.fun avg.line.fun = median

Here’s an example using a background color, and quantile lines.

pirateplot(formula = weight ~ Time, 
           data = ChickWeight,
           main = "Adding quantile lines and background colors",
           theme = 2, 
           back.col = gray(.98), # Add light gray background
           gl.col = "gray", # Gray gridlines
           gl.lwd = c(.75, 0),
           inf.f.o = .6, # Turn up inf filling
           bean.b.o = .4, # Turn down bean borders
           quant = c(.1, .9), # 10th and 90th quantiles
           quant.col = "black" # Black quantile lines
           )

Multiple IVs

You can use up to 3 categorical IVs in your plot. Here are some examples:

pirateplot(formula = height ~ sex + eyepatch + headband,
           data = pirates,
           theme = 2,
           main = "Pirate Heights")

Here’s a pirateplot with showing the relationship between movie running times based on movie genre and whether the movie is a sequel or not.

pirateplot(formula = time ~ sequel + genre + rating,
           data = subset(movies, 
                         genre %in% c("Action", "Adventure", "Comedy", "Horror") &
                         time > 0),
           main = "Movie running times",
           theme = 3,
           pal = "up")

Output

If you include the plot = FALSE argument to a pirateplot, the function will return some values associated with the plot.

times.pp <- pirateplot(formula = time ~ sequel + genre,
                       data = subset(movies, 
                         genre %in% c("Action", "Adventure", "Comedy", "Horror") &
                         time > 0),
                         plot = FALSE)

Here’s the result. The most interesting element is $summary which shows summary statistics for each bean:

times.pp
## $summary
##   sequel     genre bean.num   n       avg    inf.lb   inf.ub
## 1      0    Action        1 263 115.80228 113.79871 118.0050
## 2      1    Action        2  90 120.43333 116.09047 123.7297
## 3      0 Adventure        3 218 106.61468 103.76481 109.4544
## 4      1 Adventure        4  85 117.50588 110.70050 123.1734
## 5      0    Comedy        5 436 102.81193 101.61759 104.0678
## 6      1    Comedy        6  59 102.67797  99.73761 105.3019
## 7      0    Horror        7  82 101.92683  98.35442 105.0482
## 8      1    Horror        8  23  97.65217  93.28382 102.3052
## 
## $avg.line.fun
## [1] "mean"
## 
## $inf
## [1] "hdi"
## 
## $inf.p
## [1] 0.95

Contribute!

I am very happy to receive new contributions and suggestions to improve the pirateplot. If you come up a new theme (i.e.; customization) that you like, or have a favorite color palette that you’d like to have implemented, please contact me (yarrr.book@gmail.com) or post an issue at www.github.com/ndphillips/yarrr/issues and I might include it in a future update.

References

The pirateplot is really a knock-off of the great beanplot package and visualization from Kampstra and others (2008).

Kampstra, Peter, and others. 2008. “Beanplot: A Boxplot Alternative for Visual Comparison of Distributions.” Journal of Statistical Software 28 (1): 1–9.