how to get combined legend in ggplot2 when I have both geom_line and geom_hline

Question

I have data as follows in a file wiki-hier.dat:

    1     2     5     10    20
2   80.77 80.76 80.77 80.77 80.77
3   83.06 83.11 83.11 83.11 83.11
4   84.40 84.46 84.43 84.44 84.43
5   85.24 85.31 85.31 85.31 85.31
6   85.92 86.03 86.02 86.02 86.01
7   86.41 86.56 86.55 86.55 86.55
8   86.62 86.76 86.77 86.77 86.77
10  87.13 87.32 87.32 87.32 87.32
12  87.50 87.71 87.68 87.69 NA
14  87.72 87.93 87.94 87.94 87.94
16  87.82 88.08 88.10 88.10 NA
18  87.87 88.15 88.17 88.17 88.17
20  87.93 88.22 88.23 88.23 88.23
24  88.10 88.39 88.40 88.41 88.40
28  88.10 88.46 88.47 88.47 88.47
32  88.12 88.46 88.49 88.49 NA
36  88.16 88.50 88.52 88.52 88.52
40  88.12 88.49 88.50 88.50 88.50
50  88.11 88.48 88.45 88.44 NA
60  87.90 88.26 88.25 88.25 NA
70  87.73 88.05 88.05 88.04 NA
80  87.60 87.89 87.90 87.89 NA
100 87.38 87.66 87.63 87.62 NA
120 87.01 87.23 87.23 87.23 NA
150 86.73 86.91 86.90 86.89 NA
200 86.04 86.14 86.08 86.06 NA
250 85.62 85.59 85.53 85.51 NA

I'm trying to plot this using ggplot2. My code is as follows:

library(ggplot2)
t = read.table("wiki-hier.dat", header=TRUE)
t$x = as.numeric(rownames(t))
data = data.frame(x = rep(t$x, 2), acc = c(t$X1, t$X2),
   beam = factor(rep(c(1,2), each=length(t$x))))
pdf("plotlines-wiki-hier-ggplot.pdf")
p = qplot(x, acc, data = data, geom = c("line", "point"),
    group = beam, color = beam, shape = beam, linetype = beam) +
  xlab("K-d subdivision factor") +
  ylab("Acc@161 (pct)") +
  geom_hline(aes(yintercept=84.49, linetype="Naive Bayes"), show_guide=TRUE) +
  scale_linetype_manual("beam", values = c(1,3,2))
print(p)
dev.off()

The problem is, I get two legends, and I only want one:

enter image description here

This one legend should have colors, shapes and linetypes in it for the keys "1" and "2".

How do I fix this?

The next-best thing would be to have two legends, one containing only the keys "1" and "2" and the other containing only the key "Naive Bayes".

pressed for time at the moment, but this: http://stackoverflow.com/questions/17148679/ggplot2-need-to-construct-a-manual-legend-for-complicated-plot might help get you started constructing a custom legend. — hrbrmstr, Jun 08 '14 at 22:39

user20650 · Accepted Answer · 2014-06-09T01:00:51.937

One way is to create another dataframe which you can map the aesthetics to similar to your main data.

#Your data

dat <- structure(list(X1 = c(80.77, 83.06, 84.4, 85.24, 85.92, 86.41, 
86.62, 87.13, 87.5, 87.72, 87.82, 87.87, 87.93, 88.1, 88.1, 88.12, 
88.16, 88.12, 88.11, 87.9, 87.73, 87.6, 87.38, 87.01, 86.73, 
86.04, 85.62), X2 = c(80.76, 83.11, 84.46, 85.31, 86.03, 86.56, 
86.76, 87.32, 87.71, 87.93, 88.08, 88.15, 88.22, 88.39, 88.46, 
88.46, 88.5, 88.49, 88.48, 88.26, 88.05, 87.89, 87.66, 87.23, 
86.91, 86.14, 85.59), X5 = c(80.77, 83.11, 84.43, 85.31, 86.02, 
86.55, 86.77, 87.32, 87.68, 87.94, 88.1, 88.17, 88.23, 88.4, 
88.47, 88.49, 88.52, 88.5, 88.45, 88.25, 88.05, 87.9, 87.63, 
87.23, 86.9, 86.08, 85.53), X10 = c(80.77, 83.11, 84.44, 85.31, 
86.02, 86.55, 86.77, 87.32, 87.69, 87.94, 88.1, 88.17, 88.23, 
88.41, 88.47, 88.49, 88.52, 88.5, 88.44, 88.25, 88.04, 87.89, 
87.62, 87.23, 86.89, 86.06, 85.51), X20 = c(80.77, 83.11, 84.43, 
85.31, 86.01, 86.55, 86.77, 87.32, NA, 87.94, NA, 88.17, 88.23, 
88.4, 88.47, NA, 88.52, 88.5, NA, NA, NA, NA, NA, NA, NA, NA, 
NA)), .Names = c("X1", "X2", "X5", "X10", "X20"), class = "data.frame", 
row.names = c("2",     "3", "4", "5", "6", "7", "8", "10", "12", "14", "16", 
"18", "20", "24", "28", "32", "36", "40", "50", "60", "70", "80", "100", 
"120", "150", "200", "250"))


dat$x = as.numeric(rownames(dat))
dat = data.frame(x = rep(dat$x, 2), acc = c(dat$X1, dat$X2),
                  beam = factor(rep(c(1,2), each=length(dat$x))))

# Create a new dataframe for your horizontal line
newdf <-  data.frame(x=c(0,max(dat$x)), acc=84.49, beam='naive')

# or of you want the full horizontal lines
# newdf <-  data.frame(x=c(-Inf, Inf), acc=84.49, beam='naive')

Plot

library(ggplot2)

ggplot(dat, aes(x, acc, colour=beam, shape=beam, linetype=beam)) +
  geom_point(size=4) + 
  geom_line() +
  geom_line(data=newdf, aes(x,acc)) +
  scale_linetype_manual(values =c(1,3,2)) +
  scale_shape_manual(values =c(16,17, NA)) +
  scale_colour_manual(values =c("red", "blue", "black"))

I used NA to suppress the shape in the naive legend

enter image description here

EDIT

After re-reading perhaps all you need is this

ggplot(dat, aes(x, acc, colour=beam, shape=beam, linetype=beam)) +
  geom_point(size=4) + 
  geom_line() +
  geom_hline(aes(yintercept=84.49), linetype="dashed")

Your edit suggestion doesn't work. That's the first thing I tried, and you don't see any legend for the geom_hline(). If you explicitly add the guide, you get two legends. — Urban Vagabond, Jun 09 '14 at 04:44
@UrbanVagabond; Apologies, i misunderstood: When reading "This one legend should have colors, shapes and linetypes in it for the keys "1" and "2"." I thought that that perhaps you may just want the one legend (although prob unlikely). — user20650, Jun 09 '14 at 10:46

how to get combined legend in ggplot2 when I have both geom_line and geom_hline

1 Answers1