How to get an edge list (undirected graph) from a grouped dataframe in R?

I have created a network where people are connected to specific events (participants to event 1 connected to node "event1").

I would like to know if it is possible to remove the node "event" and directly link together the people that took part in that event.

I have done something similar in the past using Excel to work on the raw data. I would like to know if there is a faster and better way to do it without leaving R.

The dataset looks like this:

net1
from        to 
Person 1   Event1 
Person 2   Event1 
Person 3   Event2
Person 4   Event2 
Person 5   Event2 
Person 6   Event3
...

As an example, I would like to delete "Event1" and connect Person1 and Person2 together as I am doing it.

I am sorry if I cannot provide a better code to work with. But I don't know where to start from to operate in this way.

Solution

We can do this using tidyverse.

group_split with the keep = FALSE argument splits the dataframe by the to column into a list of dataframes, while removing the grouping variable in the output.
map_dfr expands each dataframe by finding all combinations of from and itself (like expand.grid). _dfr means that the output list will be row-bind to form a dataframe.
pmap_dfr operates on each row of the dataframe, and sorts horizontally (sort(c(...))). set_names is needed to line up the columns after sorting. tibble(!!! splices the sorted vector into a row of the tibble, effectively turning it into a row vector.
filter and distinct removes self-loops and duplicate links respectively.

Note that both group_split and group_map are currently experimental functions. Please use with caution.

library(tidyverse)

net1 %>%
  group_by(to) %>%
  group_split(keep = FALSE) %>%
  map_dfr(expand, crossing(from, to = from)) %>%
  pmap_dfr(~ tibble(!!!sort(c(...)) %>% set_names(c("from", "to")))) %>%
  filter(from != to) %>%
  distinct()

or with group_map:

net1 %>%
  group_by(temp = to) %>%
  group_map(~ expand(.x, crossing(from, to = from))) %>%
  ungroup() %>%
  select(-temp) %>%
  pmap_dfr(~ tibble(!!!sort(c(...)) %>% set_names(c("from", "to")))) %>%
  filter(from != to) %>%
  distinct()

or with inner_join:

net1 %>%
  inner_join(net1, by = "to") %>%
  select(from = from.x, to = from.y) %>%
  pmap_dfr(~ tibble(!!!sort(c(...)) %>% set_names(c("from", "to")))) %>%
  filter(from != to) %>%
  distinct()

We can also use graph_from_data_frame in place of pmap_dfr to return an undirected graph (be sure to load igraph before loading tidyverse, otherwise, you might get some unexpected errors):

library(igraph)
library(tidyverse)

net1 %>%
  inner_join(net1, by = "to") %>%
  select(from = from.x, to = from.y) %>%
  igraph::graph_from_data_frame(directed = FALSE) %>%
  igraph::as_data_frame(what = "edges") %>%
  filter(from != to) %>%
  distinct()

Output:

# A tibble: 4 x 2
  from     to      
  <chr>    <chr>   
1 Person_1 Person_2
2 Person_3 Person_4
3 Person_3 Person_5
4 Person_4 Person_5

Data:

net1 <- structure(list(from = c("Person_1", "Person_2", "Person_3", "Person_4", 
"Person_5", "Person_6"), to = c("Event1", "Event1", "Event2", 
"Event2", "Event2", "Event3")), class = "data.frame", row.names = c(NA, 
-6L))