Re: [igraph] Alpha centrality with valued graphs

[Brandon J. Kinne]

Dear Gabor:

Thanks very much for the quick response. I believe you're correct that
alpha centrality for valued graphs just uses the weighted adjacency
matrix. I used the function you provided, but I always receive the same
error code:

"Error in get.adjacency.dense(graph, type = type, attr = attr, names =
names,  :
 no such edge attribute"

I've specified edge weights simply as E(graph)$weight, but I always
receive the same error. As well, assigning weights to a separate R
object (for example, w <- as.vector(E(graph)$weight)) and using that as
the attributes produces the same error.

Is this a problem with attr part of the modified alpha.centrality
function, or am I just not specifying edge weights correctly?

Here's an example of my code: alph <- matrix(alpha.centrality(gr,
alpha=m, loops=F, exo=1, attr=E(gr)$weight))

where m is based on the eigenvalues of the adjacency matrix.

Thanks again for your help.

Best regards,

Brandon


Quoting Gábor Csárdi <address@hidden>:

> How exactly is 'weighted' alpha centrality defined? Is it just that we
> use the "weighted" adjacency matrix instead of the regular one. If
> yes, then I've added this to the development version. Here is the
> modified function, you can use it before the new igraph is released.
>
> Best,
> Gabor
>
> alpha.centrality <- function(graph, nodes=V(graph), alpha=1,
>                             loops=FALSE, exo=1, attr=NULL,
>                             tol=1e-7) {
>  if (!is.igraph(graph)) {
>    stop("Not a graph object")
>  }
>
>  exo <- rep(exo, length=vcount(graph))
>  exo <- matrix(exo, nc=1)
>
>  d <- t(get.adjacency(graph, attr=attr))
>  if (!loops) {
>    diag(d) <- 0
>  }
>  n <- vcount(graph)
>  id <- matrix(0, nrow=n, ncol=n)
>  diag(id) <- 1
>
>  ev <- solve(id-alpha*d, tol=tol) %*% exo
>  ev[as.numeric(nodes)+1]
> }
>
> On Sun, Jan 25, 2009 at 8:39 PM, Brandon J Kinne
> <address@hidden> wrote:
> > Dear all:
> >
> > Bonacich's 2001 paper mentions that alpha centrality can be calculated not
> > only on directed, but also valued graphs. However, I can't seem to get
> > igraph to consider edge weights when using alpha.centrality. Am I missing
> > something here, or does igraph simply not allow valued graphs with
> > alpha.centrality? I know there is other software, such as UciNet, that does
> > calculate alpha centrality for valued graphs, but I'd much prefer to stay
> > within the R environment. Thanks in advance for the help!
> >
> > Best regards,
> >
> > Brandon Kinne
> >
> >
> > _______________________________________________
> > igraph-help mailing list
> > address@hidden
> > http://lists.nongnu.org/mailman/listinfo/igraph-help
>
>
>
> --
> Gabor Csardi <address@hidden>     UNIL DGM
>
>
> _______________________________________________
> igraph-help mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/igraph-help



--
Brandon J. Kinne
Yale University
Department of Political Science
124 Prospect Street
New Haven, Connecticut 06520