CODEWITHSUNDEEP
rggplot2
Ale
Ale

Reputation: 55

Legend labels to a geom_line ggplot

Using the CPDS (Comparative Political Data Set - downloadable here), I want to plot the variables gov_left3, gov_cent3 and gov_right3 for New Zealand. For gov_left3, I'm using the colour tomato3, for gov_cent3 I use blue and for gov_right3 darkgreen. When I use ggplot to draw a geom_line, the graph is drawn perfectly, but... - I can draw the graph nicely without the legend - If I want a legend (that would be pretty nice), R is somehow mixing up the colours of gov_left3 and gov_right3 in the legend. - If i relabel the legend labels using scale_color_manual, the colours are right in the legend, but wrong in the graph.

I have tried working with several commands starting with scale, but not coming to a solution. I guess it's really simple, but I just don't see it. Do I maybe need to reshape my data frame from long/wide to wide/long?

# Import data
CPDS <- readxl::read_excel("~/Daten/CPDS.xlsx")

# Filter data
NZL <- dplyr::filter(CPDS, iso == "NZL")

# Draw geom_line -> this actually gets me the right graph and a correct legend, all I want to do now is relabel my legend
n <- ggplot() +
    geom_line(data=NZL, aes(y=gov_left3, x=year, colour="tomato3"), size=0.7, linetype = "dashed") +
    geom_line(data=NZL, aes(y=gov_cent3, x=year, colour="blue"), size=0.7, linetype = "dashed") +
    geom_line(data=NZL, aes(y=gov_right3, x=year, colour="darkgreen"), size=0.7, linetype = "dashed")

n1 <- n + geom_vline(xintercept = 1994, color="red", size=1, alpha=0.75) +
  theme_minimal() +
  labs(title = "Parliamentary seat share of all parties", subtitle = "New Zealand government, 1960 to 2014",
       x="Year", y="Seat share in %", caption = "Source: CPDS") +
  theme(plot.title = element_text(size=20, face="bold", margin = margin(10, 0, 10, 0)),
        axis.text.x = element_text(angle=45, vjust=0.5), legend.title = element_text(size=12, face="bold")) +
  scale_x_continuous(breaks = c(1960, 1970, 1980, 1990, 2000, 2010)) +
  scale_color_identity(guide="legend")

Plotting n1 gets me the graph I want, but I want to change the legend. So, instead of colour, I want Types of parties as the title. Then, for blue, I want the label centre, for tomato3 left and for darkgreen I would like right.

I hope I provided all informations needed to help :) Thank you!

enter image description here

/ Edit: According to PavoDive's help, I used the melt function to transform the data frame from wide to long. With the help of dplyr's functions filter and arrange, I created a data frame containing the three columns year, variable and value, ordered by year. So far, so good, I guess

But if I let ggplot draw a graph, the result is this What am I doing wrong?

Upvotes: 1

Views: 4842

Answers (2)

Ale
Ale

Reputation: 55

After reshaping my data from wide to long, I found out (hours later) that the column value was stored as character value, not numeric. Therefore, I had to convert the variable to numeric using the command mNZL2$value <- as.numeric(as.character(mNZL2$value)).

Plotted the whole thing like this:

ggplot(data=mNZL2, aes(x=year, y=as.numeric(value), colour=variable)) +
  geom_line(linetype="dashed", size=0.6) + 
  geom_vline(xintercept = 1996, color="red", size=1, alpha=0.75) + 
  theme_minimal() + 
  labs(title = "Parliamentary seat share of all parties", 
       subtitle = "New Zealand government, 1960 to 2016", 
       x="Year", 
       y="Seat share in %", caption = "Source: CPDS") + 
  theme(plot.title = element_text(size=20, face="bold", 
                                  margin = margin(10, 0, 10, 0)), 
        axis.text.x = element_text(angle=45, vjust=0.5), 
        legend.title = element_text(size=12, face="bold")) + 
  scale_x_continuous(breaks = c(1960, 1970, 1980, 1990, 2000, 2010)) +   
  scale_color_brewer(palette = "Accent", 
                     name = "Party types", 
                     labels=c("Right","Centre","Left")) + 
  annotate("segment", x = 2002, xend = 1997, y = 65, yend = 60,
           color = "red", size = 1, alpha = 0.75, arrow = arrow()) + 
  annotate("text", x = 2008, y = 65, label = "First MMP election",
           color = "red", fontface = "bold")

to get this:

enter image description here

Upvotes: 0

PavoDive
PavoDive

Reputation: 6496

You can solve that by setting new values in scale_color_* family, but I see an underlying problem with your approach.

It seems you are working with a wide table, instead of a long one, which is giving you unwanted problems like the one you asked about. I'll use the iris data, because you didn't provide any reproducible data:

I'll first will get a wide table with years:

dt <- iris[, 1:4] # scrapped the species variable
dt$year <- 1:150 # created a year variable

I could plot this calling each variable (Sepal-Length, Petal.Width, etc.) on an independent geom_line call. But the right thing to do is converting your data from wide to long. For that I'll use data.table::melt:

require(data.table) # library(data.table) too!
df2 <- melt(df, id.vars = "year") # check df2: is a long table now

# now the plotting part:

require(ggplot2)
ggplot(df2, aes(x = year, y = value, color = variable))+geom_line()

Now your plot has the adequate labels, and it is a simpler call. Of course you can rename the columns of df2 both at the call of melt (arguments variable.name and value.name), or directly with names(df2) <-

#### EDIT to add: ####

To change the names and values in the legend, use this at the end of your ggplot chain:

+scale_color_discrete(labels = c("center", "right", "left"), name = "Political Orientation")

Upvotes: 2

Related Questions