Adding legend to ggplot made from multiple data frames with controlled colors

Question

I have a ggplot2 line chart made from three data frames for which I have controlled the color scheme. I've instead used linetype to distinguish between lines. This leads to a situation in which a legend is not automatically generated. How can I create a legend for this plot?

tpAct <- data.frame(
  Date=seq.Date(as.Date('2017-09-01'), as.Date('2018-01-01'),by='month'),
  Reg1=rnorm(5, 10, 5),
  Reg2=rnorm(5, 15, 5),
  Reg3=rnorm(5, 20, 5),
  Reg4=rnorm(5, 25, 5),
  Reg5=rnorm(5, 30, 5),
  Total=rnorm(5, 60, 5)
)

tpOL <- data.frame(  
  Date=seq.Date(as.Date('2017-09-01'), as.Date('2018-01-01'),by='month'),
  Reg1=rnorm(5, 10, 5),
  Reg2=rnorm(5, 25, 5),
  Reg3=rnorm(5, 20, 5),
  Reg4=rnorm(5, 25, 5),
  Reg5=rnorm(5, 30, 5),
  Total=rnorm(5, 60, 5)
)

tpModL2 <- data.frame(  
  Date=seq.Date(as.Date('2017-09-01'), as.Date('2018-01-01'),by='month'),
  Reg1=rnorm(5, 10, 5),
  Reg2=rnorm(5, 25, 5),
  Reg3=rnorm(5, 20, 5),
  Reg4=rnorm(5, 25, 5),
  Reg5=rnorm(5, 30, 5),
  Total=rnorm(5, 60, 5)
)

ggplot() + 
  geom_line(data=tpAct, aes(x=Date, y=Reg1), color='red', size=1.25) +
  geom_line(data=tpAct, aes(x=Date, y=Reg2), color='blue', size=1.25) + 
  geom_line(data=tpAct, aes(x=Date, y=Reg3), color='green', size=1.25) + 
  geom_line(data=tpAct, aes(x=Date, y=Reg4), color='pink', size=1.25) + 
  geom_line(data=tpAct, aes(x=Date, y=Reg5), color='yellow', size=1.25) + 
  geom_line(data=tpAct, aes(x=Date, y=Total), color='black', size=1.25) + 
  geom_line(data=tpOL, aes(x=Date, y=Reg1), linetype=5, color='red', size=1.25) +
  geom_line(data=tpOL, aes(x=Date, y=Reg2), linetype=5, color='blue', size=1.25) +
  geom_line(data=tpOL, aes(x=Date, y=Reg3), linetype=5, color='green', size=1.25) +
  geom_line(data=tpOL, aes(x=Date, y=Reg4), linetype=5, color='pink', size=1.25) +
  geom_line(data=tpOL, aes(x=Date, y=Reg5), linetype=5, color='yellow', size=1.25) +
  geom_line(data=tpOL, aes(x=Date, y=Total), linetype=5, color='black', size=1.25) + 
  geom_line(data=tpModL2, aes(x=Date, y=Reg1), linetype=4, color='red', size=1.25) +
  geom_line(data=tpModL2, aes(x=Date, y=Reg2), linetype=4, color='blue', size=1.25) +
  geom_line(data=tpModL2, aes(x=Date, y=Reg3), linetype=4, color='green', size=1.25) +
  geom_line(data=tpModL2, aes(x=Date, y=Reg4), linetype=4, color='pink', size=1.25) +
  geom_line(data=tpModL2, aes(x=Date, y=Reg5), linetype=4, color='yellow', size=1.25) +
  geom_line(data=tpModL2, aes(x=Date, y=Total), linetype=4, color='black', size=1.25) +
  labs(x='', y='Total Balances ($B)')

I think if you properly format your data in "long" format, you can just map the data source to linetype and you'll have your legend. — Mako212
– Mako212, Commented Sep 26, 2017 at 18:57
You can do this plot with a single call to geom_line. To do that, (1) convert the individual data frames to long format, (2) stack the individual data frames into a single data frame and add an indicator column to mark the name of the source data frame. Then you can map the source data frame to linetype and Reg to color, which will give you a legend and drastically reduce the amount of code needed. If you provide samples of each of your three data frames, we can provide code to show you how to do this. — eipi10
– eipi10, Commented Sep 26, 2017 at 18:57
I appreciate the suggestion to convert the data frames to long format, but is there a way to put in a legend without doing that? That would actually take a lot more work due to how the data is set up. — Gaurav Bansal
– Gaurav Bansal, Commented Sep 26, 2017 at 19:01
It probably won't take much work. If you provide data samples, we can show you how. — eipi10
– eipi10, Commented Sep 26, 2017 at 19:02

eipi10 · Accepted Answer · 2017-09-26 19:33:55Z

3

Here's how to stack and plot the data using the sample data frames you provided:

library(tidyverse)

setNames(list(tpAct, tpOL, tpModL2), c("tpAct","tpOL","tpModL2")) %>% 
  map_df(~ .x %>% gather(key, value, -Date), .id="source") %>% 
  ggplot(aes(Date, value, colour=key, linetype=source)) +
    geom_line() +
    scale_colour_manual(values=c('red','blue','green','pink', 'yellow', 'black')) +
    theme_classic()

setNames(list(tpAct, tpOL, tpModL2), c("tpAct","tpOL","tpModL2")) puts the three data frames in a list and assigns the data frame names as the names of the list elements.

map_df(~ .x %>% gather(key, value, -Date), .id="source") converts the individual data frames to long format and stacks them into a single long-format data frame.

Here's what the plot looks like:

A faceted plot might be easier to read:

setNames(list(tpAct, tpOL, tpModL2), c("tpAct","tpOL","tpModL2")) %>% 
  map_df(~ .x %>% gather(key, value, -Date), .id="source") %>% 
  ggplot(aes(Date, value, colour=key)) +
    geom_line() +
    scale_colour_manual(values=c('red','blue','green','pink', 'yellow', 'black')) +
    theme_classic() +
    facet_grid(~ source)

edited Sep 26, 2017 at 19:33

answered Sep 26, 2017 at 19:14

eipi10

94.6k28 gold badges220 silver badges299 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Gaurav Bansal Over a year ago

Thanks. Is there a way to combine the legend into one so it says "tpActReg1" and shows as red solid line, "tpModL2Reg1" and shows a red dashed line, and so forth?

eipi10 Over a year ago

You can do this with ggplot(aes(Date, value, colour=interaction(source,key), linetype=interaction(source,key))), but the legend will have 18 entries and you'll need to do some additional work with scale_colour_manual and scale_linetype_manual to get the linetypes and colors the way you want them.

cmaimone · Accepted Answer · 2017-09-26 19:12:26Z

When you find yourself wanting to manually add a legend with ggplot2, I've found it usually means you're going about making your plot in a way other than what ggplot2 intended.

To get ggplot to generate a legend for this plot, you need to reshape your data into a long format with groups before giving it to ggplot.

You'll want to combine all three data sets (tpAct(), tpOL(), tpModL2() - I'm assuming that these are functions that are returning data frames) into a single data.frame -- let's call this combineddata. The columns would then be: dataset (to denote which set the observation is from), Date, RegType (Reg1, Reg2,..Total), and Value (the actual y values you're plotting).

Then you could create a plot with something like the following:

ggplot(combineddata, aes(x=Date, y=Value, color=RegType, linetype=dataset) + 
    geom_line()

This will draw one line for each combination of regression and data set like you have above, but will do so automatically and create a legend. If you don't like the colors and line types that ggplot picks for you, you can further specify the specific colors and line types you want to use with scales:

ggplot(combineddata, aes(x=Date, y=Value, color=RegType, linetype=dataset) + 
    geom_line() +
    scale_color_manual(values = c("red", "blue", "green", "pink", "yellow", "black")) + 
    scale_linetype_manual(values = c(1, 5, 4)

You may need to explicitly make RegType and dataset ordered factors before sending the data to ggplot to get them in the correct order, or change the order of the colors and line types above.

Collectives™ on Stack Overflow

Adding legend to ggplot made from multiple data frames with controlled colors

2 Answers 2

2 Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

2 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related