Re: [igraph] Creating subgraphs on the basis of clusters()

[Gabor Csardi]

On Fri, Dec 14, 2007 at 11:35:40AM +0900, MATSUDA, Noriyuki wrote:
> Hello:
> 
>  I need help in creating subgraphs by selecting clusters of certain sizes.  
>  To
> do so, I want to obtain the list of memberships of the vertices to be 
> selected.
> -------------------------------------------------------------------------------------------------------------------
> a <- c(0,1)
> b <- c(2,3,3,4)
> c <- c(5,6,5,7,6,7)
> d <- c(8,9,8,10,8,11, 9,10,9,11, 10,11)
> e <- c(12,13,12,15, 13,14, 14,15, 15,16)
> g<- graph(c(a,b,c,d,e),directed=F)
> plot(g, layout= layout.circle) #vertex.label=V(g)$name
> cls<- clusters(g)
> # $membership
> #  [1] 0 0 1 1 1 2 2 2 3 3 3 3 4 4 4 4 4
> # $csize
> # [1] 2 3 3 4 5   #e.g., the membership & the size of the first 
> cluster are 0 & 2
> 
> clsID<- which(cls$size == 2) - 1     # resulting in a single value 0
> which(cls$membership == clsID)  # 1, 2
> clsID<- which(cls$csize == 3) - 1  # resulting in a vector consisted of 1 & 
> 2
> which(cls$membership == clsID)  # resulting in 0 0 which are hard to 
> interpret.

clsID <- which(cls$size==n) - 1

gives you the cluster ids of the clusters with 'n' vertices, i think that
is clear. Then

which(cls$membership == clsID)

is not a very good idea, since it uses R's recycling rule and if 
clsID is longer than 1 then it is not what you want. Eg. try:

a <- 1:4
a == 1
a == c(1,2)
a == c(2,1)

I think what you want is 

which(cls$membership %in% clsID) - 1

This gives the vertex ids in the clusters clsID.

Gabor

> -------------------------------------------------------------------------------------------------------------------
> 
>   Is there a way to specify the condition like cls$membership in clsID ?
> -- 
> MATSUDA, Noriyuki　<in Kanji> 松 田 紀 之
[...]

-- 
Csardi Gabor <address@hidden>    MTA RMKI, ELTE TTK