A Comparison of plot_ly and ggplotly
for Interactive Graphs in R

Basic

We will compare these two functions basics by ploting a scatter plot to investigate the relationship between two variables: Income and Murder Rate, and then fit a linear regression line on it.

The gramma of plot_ly is similar to that of ggplot2. By specifying text section, you can customize the mouse over hover text. If text is not specified, the default hover text is just the x and y coordinates of that data point. In addtion, you can group the data points by color and size just as in ggplot2. To add a layer in the graph, such as fitting a linear regression line, we use add_trace.

fv <- state %>% lm(Murder~Income, .) %>% fitted.values()

p1 <- plot_ly(data = state, x = ~Income, y = ~Murder,
        type = "scatter", mode = "markers",
        text = ~paste(State, "<br>Income: ", Income, '<br>Murder Rate:', Murder)) %>% 
  add_trace(x = ~Income, y = fv, mode = "lines") %>%
  layout(yaxis = list(title = "Murder Rate"), showlegend = F)
p1 %>% layout(title = "Murder Rate VS Income")

To use ggplotly function, plot a regular ggplot2 graphs g and then apply ggplotly function on the plot g. The default hover text can be messy and can be rearranged easily through tooltip. In tooltip, you can sepcify which variables to be included and the order of their appearances. The rearranged hover text seems clearer than plotly’s.

g1 <- ggplot(data = state, aes(x = Income, y = Murder))+
  geom_point(aes(text = State))+
  geom_smooth(method = lm, se = FALSE)+
  labs(y = "Murder Rate")
  
ggplotly(g1 + labs(title = "Murder Rate VS Income"), 
         tooltip = c("text", "Income", "Murder"))

Subplot

To produce subplots by four different regions using plotly, we can use the subplot function which takes in a list of plots. Positions of subplots can be arrange through parameters nrows, shareX and shareY.

state %>%
  split(.$Region) %>%
  lapply(function(d) plot_ly(d, x = ~Income, y = ~Murder, color = ~Region)) %>%
  lapply(function(l) layout(l, xaxis = list(range = c(3000, 5500)), 
                            yaxis = list(range = c(1, 12)),
                            title = "Subplot: Murder Rate VS Income")) %>% 
  subplot(nrows = 2, shareX = TRUE, shareY = TRUE) 

To do it in ggplotly, useng facet_wrap in ggplot2 to create subplots and then apply ggplotly on top of that. The outcoming graph automatically has the same x axis and y axis ranges and titles for each subplot, which make it look more organized and easier to read.

g2 <- ggplot(data = state, aes(x = Income, y = Murder))+
  geom_point(aes(color = Region,text = State))+
  facet_wrap(~Region, nrow = 2)+
  labs(title = "Subplot: Murder Rate VS Income")
ggplotly(g2, tooltip = c("text", "Income", "Murder"))

Mixed Subplots

To create mixed subplots in plotly, create different plotly graphs and then use subplot function. The only problem is that we lose axis label, unless we set a location to add some words.

p3 <- plot_ly(state, x = ~reorder(Region, -Income, median), y = ~Income, 
              text = ~paste(State, "<br>Income: ", Income)) %>%
  add_boxplot(type = "box", name = "Whiskers and Outliers", boxpoints = 'outliers', showlegend = FALSE)

subplot(p3, p1) %>% layout(showlegend = FALSE,
                           annotations = list(
  list(x = 0.1 , y = 1.05, text = "1974 US Income Per Capita", showarrow = F, xref='paper', yref='paper'),
  list(x = 0.9 , y = 1.05, text = "Murder Rate VS Income", showarrow = F, xref='paper', yref='paper')))

Or we can plot regular ggplot2 graphs p1 and p2 and then use subplot(ggplotly(p1), ggplotly(p2)) In addition, there’s difference between ggplotly and ggplotly in the boxplot graph. In ggplotly, the box seems to cover the individual values, while in ggplotly, we may know the value of each points while moving the mouse onto the box. (The points is invisible because I changed the transparency.)

g3 <- ggplot(state, aes(x = reorder(Region, -Income, median), y = Income))+
  geom_point(aes(text = State), alpha = 0, size = 2)+
  geom_boxplot()
  

subplot(ggplotly(g3, tooltip = c("text", "Income")), ggplotly(g1)) %>% 
  layout(showlegend = FALSE,
         annotations = list(
  list(x = 0.1 , y = 1.05, text = "1974 US Income Per Capita", showarrow = F, xref='paper', yref='paper'),
  list(x = 0.9 , y = 1.05, text = "Murder Rate VS Income", showarrow = F, xref='paper', yref='paper')))

Multiple axes

We may see more information from a graph with multiple axes. We will compare Murder Rate with Illiteracy and Life Expection at the same time. Here, I first sort the murder rate from low to high, then the value of Illiteracy rate corresponding to left Y-axis, while life expection corresponding to the right one, in order to see the trend of these two variables as murder rate gets higher. You may click Campare data on hover on the interface to see both two values of one specific state at the same time.

Plotly provides us an option to draw multiple axes on one graph. What we need to do is to add yaxis = ‘y2’ on the second trace, and yaxis2 = list(overlaying = “y”, side = “right”) in layout. Both of two y-axes display clearly. Everything else performs good as well.

p4 <- plot_ly() %>%
  add_lines(data = state, x = ~reorder(State,Murder), y = ~Illiteracy, name = "Illiteracy Rate") %>%
  add_lines(data = state, x = ~reorder(State,Murder), y = ~Life.Exp, name = "Life Expection", 
            yaxis = 'y2') %>%
  layout(title = "Correlation between Murder Rate and Illiteracy VS Life Expection", 
         xaxis = list(title="States (Sort by Murder Rate from low to high)"),
         yaxis = list(tickfont= list(color = '#1f77b4', size=11), color='#1f77b4', range=c(0,4),
                      title="Illiteracy Rate"),
         yaxis2 = list(overlaying = "y", side = "right",
                       tickfont = list(color = '#ff7f0e', size=11), color='#ff7f0e', range=c(67.8,73.8),
                       title = "Life Expection"),
         legend = list(x = 1.05, y = 0.9))
p4

Though we are able to draw a graph with multiple axes from ggplot2 from version 2.2.0, using ggplotly to create a inference based on this graph will mess up. The reason is that, in ggplot2, adding one more axis is actually adding numeric identifiers in that specified area; it’s not a new unit of measurement. In this example, what we are doing is to linearly transform the value of life expection into the range of Illiteracy Rate. So both variables may share the same unit of measurement of default y-axis. Then we add the real value of life expection to the right y-axis. This principle also works on the second x-axis.

The following inference seems to work properly. However, there are a few drawbacks. While we drag the right y-axis, the plot of life expection doesn’t move. Instead, its movement follows the left y-axis. It’s not surprising because we know the value of life expection is using the first unit of measurement. Also, when we click Toggle Spike Lines, the Indicate Lines face toward left and down because of the same reason.

In addition, the second axis won’t show when simply adding ggplotly to a ggplot function. I guess it’s the limitation of ggplotly. I used part of the code from the previous one to solve this problem.

g4 <- ggplot(state, aes(x = reorder(State, Murder), label = Murder, label1 = Illiteracy, label2 = Life.Exp))+ 
  geom_line(aes(y = Illiteracy, colour = "Illiteracy Rate", 
                text = paste(State, '<br>Murder Rate:', Murder, '<br>Illiteracy:', Illiteracy),
                group=1))+
  geom_line(aes(y = (Life.Exp-67.8)/1.5, colour = "Life Expection",
                text = paste(State, '<br>Murder Rate:', Murder, '<br>Life.Exp:', Life.Exp), 
                group=1))+ 
  scale_y_continuous(sec.axis = sec_axis(~.*1.5+67.8, name = "Life Expection"))+ 
  labs(title = "Correlation between Murder Rate and Illiteracy VS Life Expection",
       y = "Illiteracy Rate",
       x = "States (Sort by Murder Rate from low to high)",
       colour = "Comparison")+
  theme(axis.text.x = element_text(angle = 270, hjust = 1),
        axis.text.y = element_text(colour = "#F8766D"),
        axis.title.y = element_text(colour = "#F8766D"))

ggplotly(g4,tooltip = 'text') %>%
  add_lines(data=state, x=~reorder(State, Murder), y=~Life.Exp, colors=NULL, yaxis="y2", 
             inherit=FALSE, showlegend = FALSE) %>%
  layout(yaxis2 = list(overlaying = "y", side = "right",
                       tickfont = list(color = '#00BFC4', size=11), color = '#00BFC4',
                       title = "Life Expection"),
         legend = list(x = 1.05, y = 0.95))