Joshua Rosenberg Joshua Rosenberg - 1 month ago 12
R Question

Color legend not appearing in ggnet2 plot

I am trying to plot the network

g
:

> g
Network attributes:
vertices = 866
directed = TRUE
hyper = FALSE
loops = FALSE
multiple = FALSE
bipartite = FALSE
total edges= 5310
missing edges= 0
non-missing edges= 5310

Vertex attribute names:
color degree membership vertex.names

Edge attribute names not shown


g
node
"membership" occurs with the following frequency:

> table(g %v% "membership")

1 2 3 4 5
19 44 11 196 596


When I use the excellent
ggnet2
package and the following function, a plot without a legend returns, although the documentation suggests a plot for node size, color, etc. is automatically generated (see the node legends section):

ggnet2(g_new, color = "membership", size = "degree") +
guides(size = F)


enter image description here

Fr. Fr.
Answer

My guess is that you are passing a numeric vertex attribute to ggnet2, which then understands it as a vector of color values.

Short answer:

Turn your membership attribute into a character string, and you will be fine.

Long answer:

Compare the two graphs in the example below to get the difference:

library(GGally)
library(ggplot2)
library(network)
library(sna)

g <- network(rgraph(10))
g %v% "membership" <- sample(1:5, 10, replace = TRUE)
g %v% "membership_string" <- as.character(g %v% "membership")
g %v% "degree" <- degree(g)

g
table(g %v% "membership")

## Your graph
ggnet2(g, color = "membership", size = "degree") +
  guides(size = FALSE)

## Your graph, with the membership variable as a string
ggnet2(g, color = "membership_string", size = "degree") +
  guides(size = FALSE)

If you want distinguishable colors instead of grayscale levels in your plot, you probably want something like this:

ggnet2(g, color = "membership_string", size = "degree", color.palette = "Set1") +
  guides(size = FALSE)

Also, may I recommend that you give a try to the ggnetwork package, which has additional functionalities?

Comments