4

I am trying to plot a network graph and show only the labels for the geom_node_text that have a centrality score above a specific threshold. My code is below: 

rt_tbl %>% 
mutate(centrality = centrality_authority()) %>% 
ggraph(layout = 'kk') + 
geom_edge_link(aes(), colour = "gray", alpha = 0.5) + 
geom_node_point(aes(size = centrality, colour = centrality)) + 
geom_node_text(data=subset(centrality > 0.6), aes(centrality, label=name)) + 
theme_graph(base_family = "Roboto Condensed", base_size = 13)

I encounter the following error:

Error in subset(centrality > 100) : object 'centrality' not found

My data looks like this:

# A tbl_graph: 1589 nodes and 3992 edges
# A directed multigraph with 3 components
# Node Data: 1,589 x 3 (active)
        name degree centrality
        <chr>  <dbl>      <dbl>
1   Ashinlay1970     35 0.90053429
2     WinTunMin1     25 0.66408597
3 Yaminayeyaohn1      2 0.06080755
4  TUNOO00655880      3 0.07609831
5     inewagency      8 0.21569006
6         KSwe03      4 0.12416238
# ... with 1,583 more rows

# Edge Data: 3,992 x 2
from    to
<int> <int>
1     1    48
2     1    49
3     1     1
# ... with 3,989 more rows
p0bs
  • 1,004
  • 2
  • 15
  • 22
mundos
  • 459
  • 6
  • 14
  • Your problem is not reproducible without providing at least some of your data. Is `centrality` a variable within `rt_tbl`? – Tino Nov 08 '17 at 09:19
  • Yes, centrality is a variable within rt_tbl. I have added now some data. – mundos Nov 08 '17 at 09:26

2 Answers2

10

This is not an answer but a comment for any future visitor using the answer above. When I tried levels(name) in different networks it gave erroneous results. tidygraph has filter property which can be used in various geoms for filtering nodes, edges, etc.

So, the geom_node_text line of the answer above should work better if written as:

geom_node_text(aes(filter=centrality > .6, label = name))
Alper Yilmaz
  • 319
  • 2
  • 11
  • Thanks! This is much better - levels(name) has many ways in which reading the data in can mess up the sequence of levels. – puslet88 Apr 01 '19 at 15:53
8

I never used ggraph before and you should really provide a reproducible minimal example, but try this:

rt_tbl %>% 
    mutate(centrality = centrality_authority()) %>% 
    ggraph(layout = 'kk') + 
    geom_edge_link(aes(), colour = "gray", alpha = 0.5) + 
    geom_node_point(aes(size = centrality, colour = centrality)) + 
    geom_node_text(aes(label=ifelse(centrality > .6, name, NA))) + 
    theme_graph(base_family = "Roboto Condensed", base_size = 13)

Your subset-approach does not work because it does not look inside your rt_tbl but tries to get the object centrality, which does not exist. But it would not work anyway, because you need to give it a vector of same length as your data, but subset only returns the values that match your condition. Therefore, using ifelse is better suited for your task.

Edit:

BTW this is a minimal reproducible example (at least I now know how to use ggraph):

library(tidygraph)
library(ggraph)

rt_tble <- tidygraph::create_star(10) %>% 
    mutate(centrality = centrality_authority(), 
           name = LETTERS[1:10])

ggraph(graph = rt_tble) +
    geom_edge_link() +
    geom_node_point(aes(size = centrality, colour = centrality)) + 
    geom_node_text(aes(label = ifelse(centrality > .6, as.character(name), NA_character_)))

I had to use as.character(name) instead of name or levels(name) (as I did before), maybe you would have to change that in my solution above as well...

But regarding that, have a look at @Alper Yilmaz solution below.

Tino
  • 2,091
  • 13
  • 15