Plotting two variables as lines using ggplot2 on the same graph

Question

A very newbish question, but say I have data like this:

test_data <-
  data.frame(
    var0 = 100 + c(0, cumsum(runif(49, -20, 20))),
    var1 = 150 + c(0, cumsum(runif(49, -10, 10))),
    date = seq(as.Date("2002-01-01"), by="1 month", length.out=100)
  )

How can I plot both time series var0 and var1 on the same graph, with date on the x-axis, using ggplot2? Bonus points if you make var0 and var1 different colours, and can include a legend!

I'm sure this is very simple, but I can't find any examples out there.

Plumlee

Answer 1

For a small number of variables, you can build the plot manually yourself:

ggplot(test_data, aes(date)) + 
  geom_line(aes(y = var0, colour = "var0")) + 
  geom_line(aes(y = var1, colour = "var1"))

Mcfarlane

nice example, but how to customize my own colours (E.g. black and orange)?, because it seems that you are using colour= as the variable name. – Affenpinscher

even colour='var_names' as specified by hadley works fine. but @DaveX - would be more specific if one wants to choose specific colors rather than automatically selected colours by the function. – Chose

How can I add a legend to it? – Loutish

@user1700890, the legend seems to be added automatically – Stubbed

If the colour variable is numeric, it may be necessary to as.character() on it first. – Sanborn

Answer 2

The general approach is to convert the data to long format (using melt() from package reshape or reshape2) or gather()/pivot_longer() from the tidyr package:

library("ggplot2")
library("tidyr")
library("reshape2")

## convert to long format with tidyr::pivot_longer
test_data_long_tidyr <- pivot_longer(test_data, cols = starts_with("var"))

ggplot(data=test_data_long_tidyr,
       aes(x=date, y=value, colour=name)) +
  geom_line() ## output not shown, it's equivalent to the below graph (with a tiny difference in the legend title)

## convert to long format with reshape2::melt
test_data_long <- melt(test_data, id="date")  

ggplot(data=test_data_long,
       aes(x=date, y=value, colour=variable)) +
  geom_line()

Also see this question on reshaping data from wide to long.

Daley

You can also use the gather() function of tidyr package to melt the data: gather(test_data, variable, value, -date) – Cantankerous

reshape2, and tidyr::gather are superseded (see github.com/hadley/reshape and ?tidyr::gather). Hadley recommends to use "pivot_longer" - I have added an example with the latter – Heterophyllous

Answer 3

You need the data to be in "tall" format instead of "wide" for ggplot2. "wide" means having an observation per row with each variable as a different column (like you have now). You need to convert it to a "tall" format where you have a column that tells you the name of the variable and another column that tells you the value of the variable. The process of passing from wide to tall is usually called "melting". You can use tidyr::gather to melt your data frame:

library(ggplot2)
library(tidyr)

test_data <-
  data.frame(
    var0 = 100 + c(0, cumsum(runif(49, -20, 20))),
    var1 = 150 + c(0, cumsum(runif(49, -10, 10))),
    date = seq(as.Date("2002-01-01"), by="1 month", length.out=100)
  )
test_data %>%
    gather(key,value, var0, var1) %>%
    ggplot(aes(x=date, y=value, colour=key)) +
    geom_line()

Just to be clear the data that ggplot is consuming after piping it via gather looks like this:

date        key     value
2002-01-01  var0    100.00000
2002-02-01  var0    115.16388 
...
2007-11-01  var1    114.86302
2007-12-01  var1    119.30996

Bandstand

Answer 4

I am also new to R but trying to understand how ggplot works I think I get another way to do it. I just share probably not as a complete perfect solution but to add some different points of view.

I know ggplot is made to work with dataframes better but maybe it can be also sometimes useful to know that you can directly plot two vectors without using a dataframe.

Loading data. Original date vector length is 100 while var0 and var1 have length 50 so I only plot the available data (first 50 dates).

var0 <- 100 + c(0, cumsum(runif(49, -20, 20)))
var1 <- 150 + c(0, cumsum(runif(49, -10, 10)))
date <- seq(as.Date("2002-01-01"), by="1 month", length.out=50)    

Plotting

ggplot() + geom_line(aes(x=date,y=var0),color='red') + 
           geom_line(aes(x=date,y=var1),color='blue') + 
           ylab('Values')+xlab('date')

However I was not able to add a correct legend using this format. Does anyone know how?

Fowliang

This adds a legend ggplot() + geom_line(aes(x=date,y=var0, group=1, colour = 'red')) + geom_line(aes(x=date,y=var1, group = 2, colour = 'blue')) + ylab('Values')+xlab('date') – Orthopedic

What's the difference between this and the accepted answer, besides that yours doesn't have a legend? – Aquarius

@Aquarius as indicated, the only difference is this way does not use dataframe as input, just the vectors directly – Fowliang

Answer 5

Using your data:

test_data <- data.frame(
var0 = 100 + c(0, cumsum(runif(49, -20, 20))),
var1 = 150 + c(0, cumsum(runif(49, -10, 10))),
Dates = seq.Date(as.Date("2002-01-01"), by="1 month", length.out=100))

I create a stacked version which is what ggplot() would like to work with:

stacked <- with(test_data,
                data.frame(value = c(var0, var1),
                           variable = factor(rep(c("Var0","Var1"),
                                                 each = NROW(test_data))),
                           Dates = rep(Dates, 2)))

In this case producing stacked was quite easy as we only had to do a couple of manipulations, but reshape() and the reshape and reshape2 might be useful if you have a more complex real data set to manipulate.

Once the data are in this stacked form, it only requires a simple ggplot() call to produce the plot you wanted with all the extras (one reason why higher-level plotting packages like lattice and ggplot2 are so useful):

require(ggplot2)
p <- ggplot(stacked, aes(Dates, value, colour = variable))
p + geom_line()

I'll leave it to you to tidy up the axis labels, legend title etc.

HTH

Dulin

I think you have a misplaced parens in your code up there. I think this is what you are after: stacked <- with(test_data, data.frame(value = c(var0, var1), variable = factor(rep(c("Var0", "Var1"))), each = NROW(test_data), Dates = rep(date, 2))). Also, what is the purpose of the column "each"? And is this not just a more convoluted and less efficient way to melt the data as shown by rcs? I guess I could imagine an instance where melt wouldn't get the job done, but it is almost certainly the right tool for this job unless I'm missing something? – Urina

@chase, sorry, that is Emacs ESS getting the indenting wrong. each is an argument to rep(), so we really are only getting 3 cols in stacked. I'll edit the code to make the indent clearer. – Dulin

@chase; your comment about melt() is well taken, and I note that the reshape[2] package would be useful here. I'm not that familiar with reshape2 and for such a simple manipulation doing it by hand is more complex than a call to melt(), it was less effort as I didn't need to read how to use melt(). And rcs sneaked in with his answer whilst I was producing mine; when I started the reply there had been no answers. more than one way to skin a cat - as they say! ;-) – Dulin