A first plot with Plotly

No feedback found for this session

Introduction

This is a short practical session designed to walk users through creating a first plot in plotly. You’ll need at least a bit of prior R experience before, but assuming you’ve ever written an R script before, you should be fine. The session introduces some plotly basics, and compares their implementation with similar graphs in ggplot.

At a glance…

• plotly is the major alternative to ggplot
• the code is longer, but easier to understand
• the basics are better (accessibility and interactivity, especially)
• but there’s a much smaller plotly user base
• and it’s a big jump from either base R or ggplot
• the main reason you’re not using it already: the documentation is generally poor (although Sievert’s book is very good)

Setup

library(plotly)
library(ggplot2) # for a bit of compare-and-contrast
library(dplyr) # bits of data wrangling
library(palmerpenguins) # sample data

Starting a plot object

Note the plotly is interactive from the start - mouse over, see extra information, and control e.g. zoom from the top panel menu

plotly

penguins |>
  plot_ly()

ggplot

penguins |>
  ggplot()

Adding points

Two big differences:

• plotly will make an educated guess about the best way to represent your data, while ggplot doesn’t
• plotly uses ~ to indicate how variables map onto axes etc, rather than aes()

plotly

penguins |>
  plot_ly(x = ~bill_length_mm, y = ~ bill_depth_mm) # will guess that points will work

ggplot

penguins |>
  ggplot() +
  geom_point(aes(x = bill_length_mm, y = bill_depth_mm))

Changing the type of graph

Use a type argument. Useful starter options include…

• histogram (which does statistical work for you)
• bar (which plots the supplied observations)
• scatter
• box

plotly

penguins |>
  plot_ly(x = ~ body_mass_g, 
          type = "histogram")

penguins |>
  plot_ly(y = ~ flipper_length_mm, 
          type = "box",
          name="Adding a manual label")

ggplot

penguins |>
  ggplot() +
  geom_histogram(aes(x = body_mass_g))

penguins |>
  ggplot() +
  geom_boxplot(aes(y = flipper_length_mm)) +
  xlab("Adding a manual label")

Non-spatial representations

plotly

penguins |>
  plot_ly(x = ~bill_length_mm, 
          y = ~ bill_depth_mm, 
          color = ~species, # non-UK spelling
          type = "scatter") # specify desired structure

ggplot

penguins |>
  ggplot() +
  geom_point(aes(x = bill_length_mm, 
                 y = bill_depth_mm,
                 colour = species))

Styling

plotly

penguins |>
  plot_ly(x = ~bill_length_mm, 
          y = ~ bill_depth_mm, 
          color = ~species, 
          colors = "Dark2", # using Rcolorbrewer palettes: https://cran.r-project.org/web/packages/RColorBrewer/refman/RColorBrewer.html
          type = "scatter") |> # specify desired structure
  layout(title = "100% pure plotly penguins",
         legend = list(title = list(text = "Three penguin species")))

ggplot

penguins |>
  ggplot(aes(x = bill_length_mm, 
                 y = bill_depth_mm,
                 colour = species)) + # move the aesthetics in case we decide to get fancy
  geom_point() + 
  ggtitle("Proudly pluralistic about ggplotting penguins") +
  labs(colour = "A penguin-type trio") +
  scale_colour_brewer(palette = "Dark2") # matching colorbrewer palette

Extra series

plotly

penguins |>
  plot_ly(y = ~bill_length_mm, 
          color = ~ species,
          type = "box",
          boxpoints = "all", 
          jitter = 0.3,
          pointpos = -1.8) |> # this has been the hardest example to grasp: adding a second series is idiosyncratic, and seems to depend on the type of plot you're producing in the first instance
  layout(title = "100% pure plotly penguins",
         legend = list(title = list(text = "Three penguin species")))

ggplot

penguins |>
  ggplot(aes(x = species, y = bill_length_mm, color = species)) +
  geom_boxplot() +
  # geom_jitter(width = 0.2, height = 0.1) +
  geom_point(position = position_nudge(x = -.5)) + # as far as I can tell, either nudge or jitter, but not both
  ggtitle("100% genuinely ggplotted gentoos") +
  scale_colour_discrete(name = "Three penguin species")

Bonus

plotly

penguins |>
  plot_ly(type = "splom", 
          color = ~species, 
          dimensions = list(
            list(label='flipper_length_mm', values=~flipper_length_mm),
            list(label='bill_depth_mm', values=~bill_depth_mm),
            list(label='bill_length_mm', values=~bill_length_mm),
            list(label='body_mass_g ', values=~body_mass_g )
                            ),   
          diagonal=list(visible=F))