Stacked area chart using Plotly and R without ggplot

Question

Is there any way to make a stacked bar chart only using plot_ly in R? I'm aware a possible solution is to use ggplot and then convert with ggplotly but it doesn't look as nice as other plotly charts. The Plotly site has an example, but the totals stay the same when a category is removed via clicking on the legend.

Make example data:

library(tidyverse)
library(plotly)

# Create some data
grpnames <- c("Thing_3", "Thing_2", "Thing_1")
xval <- as.factor(c(100, 101, 102, 103))
frame <- merge(grpnames, xval, all=T)
yval <- runif(12, 0, .2)
df <- tbl_df(cbind(frame, yval))
colnames(df) <- c("GroupName", "X", "Y")
df.wide <- spread(df, key = GroupName, value = Y)

The stacked bar works:

# Creates a legit stacked bar where values sum to highest point
plot_ly(df, x = ~X, y = ~Y, color = ~GroupName, type='bar') %>% 
  layout(barmode = 'stack')

I couldn't find an analogue to "barmode = 'stack'" for a line chart:

# Attempt with tidy data
df %>% 
  plot_ly(
    x = ~X, 
    y = ~Y, 
    color = ~GroupName, 
    type='scatter', 
    mode = 'lines', 
    fill = 'tonexty', 
    fillcolor = ~GroupName) 

And the example from the Plotly side, attempted here, doesn't add the values of Y for each value of X -- it simply overlays them.

# Attempt with wide data
df.wide %>% 
  plot_ly(
    x = ~X, 
    y = ~Thing_1, 
    name = 'Thing 1', 
    type = 'scatter', 
    mode = 'none', 
    fill = 'tozeroy', 
    fillcolor = 'aquamarine') %>% 
  add_trace(
    x = ~X, 
    y = ~Thing_2, 
    name = 'Thing 2', 
    fill = 'tonexty', 
    fillcolor = 'orange') %>% 
  add_trace(
    x = ~X, 
    y = ~Thing_3, 
    name = 'Thing 3', 
    fill = 'tonexty', 
    fillcolor = 'gray') 

Has anyone been able to do this successfully? Thanks!

Edit for clarification: I'm aware that it's possible to do a cumsum first and then create the chart but still appreciate the responses! I'm wondering if it's possible to do the sum within the chart so that it behaves like the stacked bars, where clicking the legend to remove a group shows the sum for the remaining groups.

Answer 1

You can adjust your data to use the cumulative sum of the y value for that point to calculate the stacked value, e.g.

library(plotly)
library(tidyverse)

       # group, sort (to keep cumulative sum in right order), and adjust Y
df %>% group_by(X) %>% arrange(GroupName) %>% mutate(Y = cumsum(Y)) %>% 
    plot_ly(type = 'scatter', x = ~X, y = ~Y, color = ~GroupName, 
            mode = 'lines', fill = 'tonexty')

Answer 2

You can calculate the heights of the stacked areas by adding together the things you want to stack. Then plot these already-stacked cumulative values. The 'reproducible' data from the original question was not reproducible, so I demonstrate with some new data here.

[note that the data used in the example on the plotly page is also converted into a cumulative table like this - https://plot.ly/r/filled-area-plots/#stacked-area-chart-with-cumulative-values]

set.seed(123)
df.wide  = data.frame(
  X = 100:105, 
  Thing_1 = cumsum(rnorm(6,10,3)), 
  Thing_2 = cumsum(rnorm(6,6,2)),
  Thing_3 = cumsum(rnorm(6,3,1)))

df.wide$T1 = df.wide$Thing_1
df.wide$T2 = df.wide$Thing_1 + df.wide$Thing_2
df.wide$T3 = df.wide$T2 + df.wide$Thing_3

  plot_ly(df.wide, fill = 'tozeroy', line = list(color = '#00000000')) %>% 
    add_trace(x = ~X, y = ~T3, name = 'Thing 3', 
              type = 'scatter', mode = 'lines',  fillcolor = 'green') %>% 
    add_trace(x = ~X, y = ~T2, name = 'Thing 2', 
              type = 'scatter', mode = 'lines',  fill = 'tozeroy', fillcolor = 'blue') %>% 
    add_trace(x = ~X, y = ~T1, name = 'Thing 1', 
              type = 'scatter', mode = 'lines',  fill = 'tozeroy', fillcolor = 'orange') 

Answer 3

For those few that stumble across this question even after a few years (like myself):

In 2021, the second link mentioned by @kkd42 is quite useful, the solution is stackgroup='one' here.

plot_ly(df,x=~X, y=~Y, color=~GroupName,
       type='scatter', mode='line', 
       stackgroup='one')

does the job for me.