R ggraph facets

Question

I'm having difficulty figuring out how to facet a graph properly.

See for example the following.

library(tidyverse)
library(tidygraph)
library(ggraph)

myedges <- data.frame(from=rep(letters[1:5],2),
                      to=c(rep(LETTERS[12:13],2),
                      rep(LETTERS[14],3),
                      LETTERS[15:16], 
                      LETTERS[17]),
                      n_user=sample(1:20, 10, replace=TRUE),
                      f=rep(c("face_A", "face_B"),5)
                      ) 


mynodes <- myedges %>% 
  pivot_longer(cols = c(from, to),
               names_to = "od",
               values_to = "node_name") %>% 
  group_by(node_name) %>% 
  summarise(n_user=sum(n_user)) 


g<-tbl_graph(nodes = mynodes, edges= myedges)

# Arc diagram with edge faceting, it seems ok but the nodes
ggraph(g, layout = 'linear') + 
  geom_edge_arc(aes(edge_width=n_user, label=n_user), label_pos=0.1, edge_alpha=0.2)+
  geom_node_point(aes(size=n_user), alpha=0.5, show.legend = FALSE)+
  geom_node_text(aes(label = node_name), size=3, nudge_y = - 0.5)+
  facet_edges(facets = vars(f), scales = "free")

The graph is almost fine except that I would like to have for each panel just the nodes that are effectively connected by edges: i.e. not having "orphan" nodes like the node "O" in the panel "face_A" or the node "P" in the panel "face_B".

Given that, I expect facet_nodes would be my potential solution by drawing just the nodes mapped by a variable and the corresponding edges for the nodes present in that panel.

I expect to accomplish that I need to introduce the mapping variable in nodes but here comes the problem because I can not conceive a proper way to do that.

The expected result is a graph with each panel node connected by edges

Any suggestions for that?

score 1 · Accepted Answer · answered Dec 04 '22 at 15:28

I think the problem here is that you are trying to plot two subgraphs using facets, but the nodes are not mutually exclusive across the faceting variable. This means you can't use facet_node. Since you are faceting by an edge attribute, you need to use facet_edge, but unfortunately facet_edge plots all the nodes, with no option to remove isolated nodes.

The obvious way to get round this (even though it's a bit laborious) is to plot two different subgraphs and join them with patchwork:

library(patchwork)

gA <- g %>% 
  activate(edges) %>% 
  filter(f == "face_A") %>% 
  activate(nodes) %>% 
  filter(!node_is_isolated())

gB <- g %>% 
  activate(edges) %>% 
  filter(f == "face_B") %>% 
  activate(nodes) %>% 
  filter(!node_is_isolated())

p1 <- ggraph(gA, layout = 'linear') + 
  geom_edge_arc(aes(edge_width = n_user, label = n_user), 
                label_pos = 0.1, edge_alpha = 0.2) +
  geom_node_point(aes(size=n_user), alpha = 0.5, show.legend = FALSE) +
  geom_node_text(aes(label = node_name), size = 3, nudge_y = - 0.5)  +
  facet_grid(.~f) +
  theme(legend.position = "none")

p2 <- ggraph(gB, layout = 'linear') + 
  geom_edge_arc(aes(edge_width = n_user, label = n_user), 
                label_pos = 0.1, edge_alpha = 0.2) +
  geom_node_point(aes(size = n_user), alpha = 0.5, show.legend = FALSE) +
  geom_node_text(aes(label = node_name), size = 3, nudge_y = - 0.5) +
  facet_grid(.~f)

p1 + p2

yes, thank you! I had also thought about the filtering by facets but in my original intent that was "before" the creation of the graph; I was wondering about the storing of the graphs in a list by means of a loop to be traversed along the number of facets; I'm still searching for a viable solution in that direction although... — maxbre, Dec 04 '22 at 17:51
I need o keep the solution enough general because of the number of facets in my real dataset and also to keep the code as much compact as possible — maxbre, Dec 04 '22 at 19:14

score 1 · Answer 2 · answered Dec 04 '22 at 20:57

by further considering the above approach I came out with this sort of solution that still needs to be rounded off in some corners; some passages of the code are rather rough but the overall solution deserves to point in the direction I would like to go further

I'm posting here the complete code

library(tidyverse)
library(tidygraph)
library(ggraph)

myedges <- data.frame(from=rep(letters[1:5],2),
                      to=c(rep(LETTERS[12:13],2),
                           rep(LETTERS[14],3),
                           LETTERS[15:16], 
                           LETTERS[17]),
                      n_user=sample(1:20, 10, replace=TRUE),
                      f=rep(c("face_A", "face_B"),5)
) 


mynodes <- myedges %>% 
  pivot_longer(cols = c(from, to),
               names_to = "od",
               values_to = "node_name") %>% 
  group_by(node_name) %>% 
  summarise(n_user=sum(n_user)) 

g<-tbl_graph(nodes = mynodes, edges= myedges)


graph_list <- list() 
for(i in unique(myedges$f)) {
  
  gr <- g %>% 
    activate(edges) %>% 
    filter(f == i) %>% 
    activate(nodes) %>% 
    filter(!node_is_isolated())  
  
  graph_list[[i]] <- gr
}

plot_list <- list()
for(i in seq_along(graph_list)){
  
  pl<-ggraph(graph_list[[i]], layout = 'linear') + 
    geom_edge_arc(aes(edge_width = n_user, label = n_user), 
                  label_pos = 0.1, edge_alpha = 0.2) +
    geom_node_point(aes(size=n_user), alpha = 0.5, show.legend = FALSE) +
    geom_node_text(aes(label = node_name), size = 3, nudge_y = - 0.5)  +
    facet_grid(.~f) +
    theme(legend.position = "none")
  
  plot_list[[i]] <- pl
} 

library(ggpubr)

ggarrange(plotlist = plot_list, 
          ncol = max(seq_along(graph_list))
)

thank you for the hints

R ggraph facets

2 Answers2