igraph从邻接表生成邻接矩阵

Question

I have friendship data in an adjacency list. Every person in the sample (denoted by id) could nominate up to 5 friends (f1-f5). The data frame called "net_test" looks like this:

    id   f1   f2   f3   f4   f5
1  1101 1113 1112   NA   NA   NA
2  1102 1111 1113 1103 1105   NA
3  1103 1105 1110   NA   NA   NA
4  1104 1115 1106 1110 1109 1112
5  1105 1103 1109 1116 1101   NA
6  1106 1121 1103 1113   NA   NA
7  1107 1106 1111   NA   NA   NA
8  1108 1104 1109   NA   NA   NA
9  1109 1114 1103 1113 1108 1120
10 1110 1101 1103 1109 1107   NA

The first row is the column number. From this data I'd like to generate an adjacency matrix with the ids as row and column names, and entries 1 if two ids have a friendship link. I tried saving it as a graph first, and then generate the adjacency matrix in a second step:

require(igraph)
netdat<-graph.adjlist(net_test, mode="out", duplicate=FALSE)
adjmat <- get.adjacency(netdat, type="both")

When I apply the graph.adjlist command, the following error occurs:

At structure_generators.c:84 : Invalid (negative) vertex id, Invalid vertex id

Would there be another way to convert the adjacency list into an adjacency matrix?

Answer 1

You misunderstood the commands. The command get.adjlist takes as parameter an igraph graph object, and returns a list-type object representation of the graph. You are applying this to a data frame which is not being coerced to an igraph object.

Below is the correct way to construct a igraph graph object using a data frame, and how to get various graph representations of this object.

require(reshape2)
net_list <- melt( net_test, id.vars = "id")
net_list <- net_list[ !is.na(net_list$value), c("id", "value") ]
graph_o <- graph.data.frame(net_list) #This is a proper igraph graph object
#got from a data frame directly

list_rep <- get.adjlist(graph_o) #this now returns an adjacency list 
#representation of your graph
matrix_rep <- get.adjacency(graph_o) #this gives you the adjacency
#matrix as a (sparse) matrix with the row and column names as you want.

Answer 2

I guess the problem is, that net_test is no proper adjacency list, because it contains more than just two columns. So the first thing I'd do is bringing it to this form by melting:

require(reshape2)
net_list <- melt( net_test, id.vars = "id")
net_list <- net_list[ !is.na(net_list$value), c("id", "value") ]
colnames(net_list) <- c("from", "to")
graph.adjlist(net_list, mode="out", duplicate=FALSE)

# IGRAPH D--- 1121 66 --

If you just need the adjacency matrix and without further need for igraph, you could directly cast the molten net_list

acast(net_list, from ~ to, fun.aggregate = length )