Introduction to Plotly
Andy Grogan-Kaylor
September 02, 2021
Background
There are an ever increasing number of animated, or interactive, graphical libraries for R. Choosing which library to use, and which library will be persistent and stable over time is a challenge.
plotly is a JavaScript graphing library that creates graphics for the web.
plotly for R is one of the easiest, and most richly featured libraries to use for creating graphics on the web.
install.packages("plotly")
The Best Place to Start With Plotly Is ggplotly()
library(ggplot2) # grammar of graphics
library(ggthemes) # beautiful additional themes and colors
library(plotly) # interactive graphs
library(dplyr) # data wrangling
load("WorldBankData.RData") # load the data
Construct a Simple ggplot
Remember all ggplots start simple, and then are tweaked…
# NB, I am mapping the plot into an object
# this is a tremendously useful idea, and will come in handy...
p0 <- ggplot(mydata, # the data I am using
aes(x = GDP, # how I am graphing it
y = life_expectancy, # NB I'm 'doubling up'
color = life_expectancy)) + # on dimensions
geom_point() + # the geometries I want to use to graph
geom_smooth()
# this graph needs better labels and titles
p0 # I am replaying the object, i.e. the plot
Construct a More Complicated ggplot
# NB, I am mapping the plot into an object
# this is a tremendously useful idea, and will come in handy...
p1 <- ggplot(mydata, # the data I am using
aes(x = GDP, # how I am graphing it
y = life_expectancy, # NB I'm 'doubling up'
color = life_expectancy)) + # on dimensions
scale_color_gradient(low = "white",
high = "red") + # CONTINUOUS palette
geom_point() + # the geometries I want to use to graph
geom_smooth()
# this graph needs better labels and titles
p1 # I am replaying the object, i.e. the plot
plotly-etize it
Mouse over the plot to see the interactivity.
Piping
Pipes %>% connect pieces of a command e.g. data to a graph command.
# just to make sure we have %>%'s
library(dplyr) # data wrangling
Use dplyr to remove “Aggregates”
mydata <- mydata %>% filter(region != "Aggregates")
Use plotly syntax
Plotly Has Similar Logic to ggplot.
# mapping plot into an object
# to create the graph, connecting 2 components with a pipe
p2 <- mydata %>% # pipes connect the pieces
plot_ly(x = ~GDP, # dimensions of the data we are using
y = ~life_expectancy, # NB the '~'
color = ~life_expectancy, # double counting a dimension
colors = c("red", "blue")) # specifying color
p2 # replay the graph
3D Scatterplot
plotly does not have every geometry that ggplot2 has, and vice versa.
Most notably, a smoother geometry seems to be missing.
Note also, the cryptic (to me) use of (embedded) lists to style the graph.
A 3D Graph!
p3 <- mydata %>% # pipe connects data to plot
plot_ly(x = ~GDP, # dimensions of the data we are using
z = ~life_expectancy, # NB the '~'
y = ~Gini,
color = ~life_expectancy,
colors = c("red", "blue"),
text = ~country_name) %>% # pipe connects to layout
layout(title = 'Life Expectancy by GDP and Gini',
scene = list(xaxis = list(title = 'GDP'),
yaxis = list(title = 'Gini Inequality Coefficient'),
zaxis = list(title = 'Life Expectancy')))
p3 # replay the graph
Let’s Think About the Idea of Dimension
On the previous 3D graphs, we had 3 dimensions of GDP, life expectancy and Gini coefficient of inequality.
Do the preceding 3D graph, or the 2D graph on the next slide do a better job of representing multiple dimensions? What are the best ways of representing additional dimensions of data?
- A third spatial dimension?
- Color? Shape?
- Facet wrapping (small multiples?)
A 2D Version With Some of the Same Style Elements
p4 <- mydata %>% # pipe connects data to plot
plot_ly(x = ~GDP, # dimensions of the data we are using
y = ~life_expectancy, # NB the '~'
color = ~life_expectancy,
colors = c("red", "blue"),
text = ~country_name) %>% # pipe connects to layout
layout(title = 'Life Expectancy by GDP',
xaxis = list(title = 'GDP'),
yaxis = list(title = 'Life Expectancy'))
p4 # replay the graph
A Bar Graph!
Because plotly does not have some features of ggplot, we need to use dplyr to manage our data (summarize scores) before sending to plotly.
mydata %>% # use mydata
group_by(region) %>% # group by region
filter(!is.na(life_expectancy)) %>% # filter on non-missing
summarise(meanLIFE = mean(life_expectancy)) %>% # get averages
plot_ly(x = ~region, # x is region
y = ~meanLIFE) # y is MEAN life expectancy
A Map!!
plotly relies on ISO standard codes for country names in order to make maps. plotly can also use State abbreviations, and latitudes and longitudes.
mymap <- mydata %>%
filter(year == 2015) %>% # use dplyr to filter on year
plot_geo(locations = ~iso3c, # ISO3 Country code
color = ~life_expectancy,
z = ~life_expectancy,
text = ~country_name) %>%
layout(title = "Countries by Life Expectancy in 2015",
geo = list(showland = FALSE,
showcountries = TRUE)) %>%
colorbar(title = 'life expectancy')
mymap # replay mymap
A Globe!!!
myglobe <- mydata %>%
filter(year == 2015) %>%
plot_geo(locations = ~iso3c,
color = ~life_expectancy,
z = ~life_expectancy,
text = ~country_name) %>%
layout(title = "Countries by Life Expectancy in 2015",
geo = list(showland = FALSE,
showcountries = TRUE,
projection = list(type = 'orthographic',
rotation = list(lon = -30,
lat = 10,
roll = 0)))) %>%
colorbar(title = 'life expectancy')
myglobe
Questions?