For Social Network Analysis-Graph Analysis - How to convert 2 mode data to 1 mode data?

[messings]

Hi,

 

Does anyone know of a package in R that has a function to convert
network data (e.g. an adjacency matrix or ) from 2-mode to 1-mode?  I am
conducting social network analysis.  I know that Pajek has this function
under Net --> Transform --> 2-mode to 1-mode --> Rows.  I have searched
the documentation under packages 'sna', 'network', 'igraph', and
'dynamicgraph' but I was not able to identify a comparable function.  

 

I would just export my data to Pajek and import it to R, but I'm going
to have to generate hundreds of these graphs, so it would take quite a
bit of time to do it this way.  

 

Thanks,

 

Solomon

 


        [[alternative HTML version deleted]]

______________________________________________
R-help@... mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

[Gabor Csardi]

Solomon, if i understand two-mode networks properly (they're bipartite, right?),
then this is not hard to do with igraph. Basically, for each vertex create an
order=2 neighborhood, and then create a graph from the adjacency list,
it is something like this:

two.to.one <- function(g, keep) {
  neis <- neighborhood(g, order=2)
  neis <- lapply(seq(neis), function(x) neis[[x]][ neis[[x]] != x-1]) ## drop self-loops
  neis <- lapply(neis, function(x) x[ x %in% keep ])                  ## keep only these
  neis <- lapply(seq(neis), function(x) t(cbind(x-1, neis[[x]])))     ## create edge lists
  neis[-keep-1] <- NULL                                               ## these are not needed
  neis <- matrix(unlist(neis), byrow=TRUE, nc=2)                      ## a single edge list
  neis <- neis[ neis[,1] > neis[,2], ]                                ## count an edge once only
  mode(neis) <- "character"
  g2 <- graph.edgelist(neis, dir=FALSE)
  V(g2)$id <- V(g2)$name                                              ## 'id' is used in Pajek
  g2
}

It does not check that the graph is indeed two-mode, and it keeps the
vertices given in the 'keep' argument. 'keep' is made of igraph vertex ids.
You can use it like this:

g <- graph.ring(10)
keep <- seq(0,8,by=2) ## we keep the 'even' vertices

g2 <- two.to.one(g, keep)
write.graph(two.to.one(g, keep), format="pajek", file="/tmp/a.net")

I haven't tested it much. We'll have a better function in the next igraph version.
Gabor

On Fri, May 09, 2008 at 03:37:05PM -0400, Messing, Solomon O. wrote:

> Hi,
>
>  
>
> Does anyone know of a package in R that has a function to convert
> network data (e.g. an adjacency matrix or ) from 2-mode to 1-mode?  I am
> conducting social network analysis.  I know that Pajek has this function
> under Net --> Transform --> 2-mode to 1-mode --> Rows.  I have searched
> the documentation under packages 'sna', 'network', 'igraph', and
> 'dynamicgraph' but I was not able to identify a comparable function.  
>
>  
>
> I would just export my data to Pajek and import it to R, but I'm going
> to have to generate hundreds of these graphs, so it would take quite a
> bit of time to do it this way.  
>
>  
>
> Thanks,
>
>  
>
> Solomon
>
>  
>
>
> [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help@... mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

--
Csardi Gabor <csardi@...>    UNIL DGM

______________________________________________
R-help@... mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

[messings]

Hi Gabor,

Thank you for your help, and thanks for making the excellent igraph
package.  The function below seems not generate an edge list that works
for my data.  I coerced a my data from a data frame using
graph.data.frame.  

You asked in your previous post if 2-mode networks are bipartite.  I
believe the answer is yes.  However, in

Hanneman, Robert A. and Mark Riddle.  2005.  Introduction to social
network methods.  Riverside, CA:  University of California, Riverside (
published in digital form at http://faculty.ucr.edu/~hanneman/ )

I found the following:

"Two-mode data are sometimes stored in a second way, called the
"bipartite" matrix.  A bipartite matrix is formed by adding the rows as
additional columns, and columns as additional rows."

Did I need to convert my data frame to a bipartite matrix before
applying the two.to.one function?

Solomon


>-----Original Message-----
>From: Gabor Csardi [mailto:csardi@...]
>Sent: Saturday, May 10, 2008 1:09 PM
>To: Messing, Solomon O.
>Cc: r-help@...
>Subject: Re: [R] For Social Network Analysis-Graph Analysis - How to
convert 2
>mode data to 1 mode data?
>
>Solomon, if i understand two-mode networks properly (they're bipartite,
>right?),
>then this is not hard to do with igraph. Basically, for each vertex
create an
>order=2 neighborhood, and then create a graph from the adjacency list,
>it is something like this:
>
>two.to.one <- function(g, keep) {
>  neis <- neighborhood(g, order=2)
>  neis <- lapply(seq(neis), function(x) neis[[x]][ neis[[x]] != x-1])
## drop
>self-loops
>  neis <- lapply(neis, function(x) x[ x %in% keep ])
## keep
>only these
>  neis <- lapply(seq(neis), function(x) t(cbind(x-1, neis[[x]])))
## create
>edge lists
>  neis[-keep-1] <- NULL
## these
>are not needed
>  neis <- matrix(unlist(neis), byrow=TRUE, nc=2)
## a
>single edge list
>  neis <- neis[ neis[,1] > neis[,2], ]
## count
>an edge once only
>  mode(neis) <- "character"
>  g2 <- graph.edgelist(neis, dir=FALSE)
>  V(g2)$id <- V(g2)$name
## 'id'
>is used in Pajek
>  g2
>}
>
>It does not check that the graph is indeed two-mode, and it keeps the
>vertices given in the 'keep' argument. 'keep' is made of igraph vertex
ids.
>You can use it like this:
>
>g <- graph.ring(10)
>keep <- seq(0,8,by=2) ## we keep the 'even' vertices
>
>g2 <- two.to.one(g, keep)
>write.graph(two.to.one(g, keep), format="pajek", file="/tmp/a.net")
>
>I haven't tested it much. We'll have a better function in the next
igraph

>version.
>Gabor
>
>On Fri, May 09, 2008 at 03:37:05PM -0400, Messing, Solomon O. wrote:
>> Hi,
>>
>>
>>
>> Does anyone know of a package in R that has a function to convert
>> network data (e.g. an adjacency matrix or ) from 2-mode to 1-mode?  I

am
>> conducting social network analysis.  I know that Pajek has this
function
>> under Net --> Transform --> 2-mode to 1-mode --> Rows.  I have
searched
>> the documentation under packages 'sna', 'network', 'igraph', and
>> 'dynamicgraph' but I was not able to identify a comparable function.
>>
>>
>>
>> I would just export my data to Pajek and import it to R, but I'm
going
>> to have to generate hundreds of these graphs, so it would take quite
a

>> bit of time to do it this way.
>>
>>
>>
>> Thanks,
>>
>>
>>
>> Solomon
>>
>>
>>
>>
>> [[alternative HTML version deleted]]
>>
>> ______________________________________________
>> R-help@... mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide

http://www.R-project.org/posting-guide.html
>> and provide commented, minimal, self-contained, reproducible code.
>
>--
>Csardi Gabor <csardi@...>    UNIL DGM

______________________________________________
R-help@... mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

[Gabor Csardi]

Hi Solomon,

On Tue, May 13, 2008 at 05:53:44PM -0400, Messing, Solomon O. wrote:

> Hi Gabor,
>
> Thank you for your help, and thanks for making the excellent igraph
> package.  The function below seems not generate an edge list that works
> for my data.  I coerced a my data from a data frame using
> graph.data.frame.  
>
> You asked in your previous post if 2-mode networks are bipartite.  I
> believe the answer is yes.  However, in
>
> Hanneman, Robert A. and Mark Riddle.  2005.  Introduction to social
> network methods.  Riverside, CA:  University of California, Riverside (
> published in digital form at http://faculty.ucr.edu/~hanneman/ )
>
> I found the following:
>
> "Two-mode data are sometimes stored in a second way, called the
> "bipartite" matrix.  A bipartite matrix is formed by adding the rows as
> additional columns, and columns as additional rows."
>
> Did I need to convert my data frame to a bipartite matrix before
> applying the two.to.one function?

No. The function i've sent assumes that your network is bipartite, i.e.
if A and B are connected by an edge, then they're assumed to be
different types of nodes. Just create the graph, calculate the 'keep'
parameter, I assume that you know this from external information, and
then call the function.

G.

> Solomon
>
>
> >-----Original Message-----
> >From: Gabor Csardi [mailto:csardi@...]
> >Sent: Saturday, May 10, 2008 1:09 PM
> >To: Messing, Solomon O.
> >Cc: r-help@...
> >Subject: Re: [R] For Social Network Analysis-Graph Analysis - How to
> convert 2
> >mode data to 1 mode data?
> >
> >Solomon, if i understand two-mode networks properly (they're bipartite,
> >right?),
> >then this is not hard to do with igraph. Basically, for each vertex
> create an
> >order=2 neighborhood, and then create a graph from the adjacency list,
> >it is something like this:
> >
> >two.to.one <- function(g, keep) {
> >  neis <- neighborhood(g, order=2)
> >  neis <- lapply(seq(neis), function(x) neis[[x]][ neis[[x]] != x-1])
> ## drop
> >self-loops
> >  neis <- lapply(neis, function(x) x[ x %in% keep ])
> ## keep
> >only these
> >  neis <- lapply(seq(neis), function(x) t(cbind(x-1, neis[[x]])))
> ## create
> >edge lists
> >  neis[-keep-1] <- NULL
> ## these
> >are not needed
> >  neis <- matrix(unlist(neis), byrow=TRUE, nc=2)
> ## a
> >single edge list
> >  neis <- neis[ neis[,1] > neis[,2], ]
> ## count
> >an edge once only
> >  mode(neis) <- "character"
> >  g2 <- graph.edgelist(neis, dir=FALSE)
> >  V(g2)$id <- V(g2)$name
> ## 'id'
> >is used in Pajek
> >  g2
> >}
> >
> >It does not check that the graph is indeed two-mode, and it keeps the
> >vertices given in the 'keep' argument. 'keep' is made of igraph vertex
> ids.
> >You can use it like this:
> >
> >g <- graph.ring(10)
> >keep <- seq(0,8,by=2) ## we keep the 'even' vertices
> >
> >g2 <- two.to.one(g, keep)
> >write.graph(two.to.one(g, keep), format="pajek", file="/tmp/a.net")
> >
> >I haven't tested it much. We'll have a better function in the next
> igraph
> >version.
> >Gabor
> >
> >On Fri, May 09, 2008 at 03:37:05PM -0400, Messing, Solomon O. wrote:
> >> Hi,
> >>
> >>
> >>
> >> Does anyone know of a package in R that has a function to convert
> >> network data (e.g. an adjacency matrix or ) from 2-mode to 1-mode?  I
> am
> >> conducting social network analysis.  I know that Pajek has this
> function
> >> under Net --> Transform --> 2-mode to 1-mode --> Rows.  I have
> searched
> >> the documentation under packages 'sna', 'network', 'igraph', and
> >> 'dynamicgraph' but I was not able to identify a comparable function.
> >>
> >>
> >>
> >> I would just export my data to Pajek and import it to R, but I'm
> going
> >> to have to generate hundreds of these graphs, so it would take quite
> a
> >> bit of time to do it this way.
> >>
> >>
> >>
> >> Thanks,
> >>
> >>
> >>
> >> Solomon
> >>
> >>
> >>
> >>
> >> [[alternative HTML version deleted]]
> >>
> >> ______________________________________________
> >> R-help@... mailing list
> >> https://stat.ethz.ch/mailman/listinfo/r-help
> >> PLEASE do read the posting guide
> http://www.R-project.org/posting-guide.html
> >> and provide commented, minimal, self-contained, reproducible code.
> >
> >--
> >Csardi Gabor <csardi@...>    UNIL DGM

--
Csardi Gabor <csardi@...>    UNIL DGM

______________________________________________
R-help@... mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

[Gabor Csardi]

Please stay on the list.

On Tue, May 13, 2008 at 06:05:15PM -0400, Messing, Solomon O. wrote:
> Gabor,
>
> By the way, this seems to work:

I'm a bit lost. So now you're converting your data frame
to a matrix? Why? Or you're doing the two-mode to one-mode
conversion here? It does not seem so to me.

Btw. there is a get.adjacency function in igraph to convert
a graph to an adjacency matrix.

G.

>
> df.to.nxn <- function( x, y ){
> # x values will be the N x N values  
>     M <- matrix( nrow = length( unique( x ) ), ncol = length( unique( x
> ) ),
>           dimnames = list( unique( x ), unique( x ) ) )
>     M[ 1:length( unique( x ) ), 1:length( unique( x ) ) ] <- 0
> # initialize the values to 0
>     for( i in 1:length( x ) ) {
> # iterate through rows of data    
>             index = which( y == y[i] )                    
>                 M[ as.character( x[ index ] ), as.character( x[ index ]
> ) ] = 1        
>      }
> M
> # return M, an N x N matrix
> }

--
Csardi Gabor <csardi@...>    UNIL DGM

______________________________________________
R-help@... mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

[messings]

Consider the following two mode-data:
 
edgelist:
  actor event
1   Sam     a
2   Sam     b
3   Sam     c
4  Greg     a
5   Tom     b
6   Tom     c
7   Tom     d
8  Mary     b
9  Mary     d
 
Two-Mode Adjacency Matrix:
     a b c d
Sam  1 1 1 0
Greg 1 0 0 0
Tom  0 1 1 1
Mary 0 1 0 1
 
To transform two mode to one mode data, we need a function that transforms the data like so:
 
Sam is connected to Greg (via event a)
Sam is connected to Tom (via event b and c)
Sam is connected to Mary (via event b)
Tom is connected to Mary (via event b and d)
 
OK, now I load my data by executing the following:
###################################################################################
require(igraph)
df <- data.frame(actor = c('Sam','Sam','Sam','Greg','Tom','Tom','Tom','Mary','Mary'),
                  event =c('a','b','c','a','b','c','d','b','d') )
g = graph.data.frame(df, directed=F)  #Coerce data to igraph object 'g'

#Loading Function two.to.one:
##two.to.one() transforms 2-mode data to 1-mode
two.to.one <- function(g, keep) {
 neis <- neighborhood(g, order=2)
 neis <- lapply(seq(neis), function(x) neis[[x]][ neis[[x]] != x-1]) ## drop self-loops
 neis <- lapply(neis, function(x) x[ x %in% keep ])                  ## keep only these
 neis <- lapply(seq(neis), function(x) t(cbind(x-1, neis[[x]])))     ## create edge lists
 neis[-keep-1] <- NULL                                               ## these are not needed
 neis <- matrix(unlist(neis), byrow=TRUE, nc=2)                      ## a single edge list
 neis <- neis[ neis[,1] > neis[,2], ]                                ## count an edge once only
 mode(neis) <- "character"
 g2 <- graph.edgelist(neis, dir=FALSE)
 V(g2)$id <- V(g2)$name          ## 'id' is used in Pajek
 g2
}

#Actors are the first 4 verticies, set them to be kept:
keep = V(g)[1:4]                
#Convert matrix with two.to.one:
g2 = two.to.one(g, keep)
g2
###################################################################################
This yields the following output:
> g2
Vertices: 4
Edges: 2
Directed: FALSE
Edges:
         
[0] 3 -- 2
[1] 4 -- 1

But, this can't be right.  Here there are only two edges where there should be four, and if I am inturpreting correctly, the output it is reporting that Tom is connected to Greg (he is not) and Sam is connected to Mary (which is true).

When I load my function, which is designed to transform a two mode edgelist (e.g. two columns of data) into a one-mode adjacency matrix it seems to work:
###################################################################################
#load my function
df.to.nxn <- function( x, y ){                                                     # x values will be the N x N values  
    M <- matrix( nrow = length( unique( x ) ), ncol = length( unique( x ) ),
          dimnames = list( unique( x ), unique( x ) ) )
    M[ 1:length( unique( x ) ), 1:length( unique( x ) ) ] <- 0                    #initialize the values to 0 - this possibly could be removed for illustrative purposes    
    for( i in 1:length( x ) ) {                                                   # iterate through rows of data    
            index = which( y == y[i] )                    
                M[ as.character( x[ index ] ), as.character( x[ index ] ) ] = 1        
     }
M                                                                                 # return M, an N x N matrix
}
#Convert matrix
g3 = df.to.nxn(df$actor, df$event)
g4 = graph.adjacency(g3, mode = "undirected", diag = F)
V(g4)$name = row.names(g3)
g4
###################################################################################
This yields:
> g4
Vertices: 4
Edges: 4
Directed: FALSE
Edges:
               
[0] Sam  -- Greg
[1] Sam  -- Tom
[2] Sam  -- Mary
[3] Tom  -- Mary

Which is what we wanted.  I have not figured out how to weight edges yet (the Sam and Tom edge and the Tom and Mary edge should perhaps be weighted at 2 because 'connected twice' -- connected by two events).

-Solomon

________________________________

From: Gabor Csardi [mailto:csardi@...]
Sent: Wed 5/14/2008 4:01 AM
To: Messing, Solomon O.
Cc: R Help list
Subject: Re: [R] For Social Network Analysis-Graph Analysis - How to convert 2 mode data to 1 mode data?



Please stay on the list.

On Tue, May 13, 2008 at 06:05:15PM -0400, Messing, Solomon O. wrote:
> Gabor,
>
> By the way, this seems to work:

I'm a bit lost. So now you're converting your data frame
to a matrix? Why? Or you're doing the two-mode to one-mode
conversion here? It does not seem so to me.

Btw. there is a get.adjacency function in igraph to convert
a graph to an adjacency matrix.

G.

>
> df.to.nxn <- function( x, y ){
> # x values will be the N x N values  
>     M <- matrix( nrow = length( unique( x ) ), ncol = length( unique( x
> ) ),
>           dimnames = list( unique( x ), unique( x ) ) )
>     M[ 1:length( unique( x ) ), 1:length( unique( x ) ) ] <- 0
> # initialize the values to 0
>     for( i in 1:length( x ) ) {
> # iterate through rows of data  
>             index = which( y == y[i] )                  
>                 M[ as.character( x[ index ] ), as.character( x[ index ]
> ) ] = 1      
>      }
> M
> # return M, an N x N matrix
> }

--
Csardi Gabor <csardi@...>    UNIL DGM



        [[alternative HTML version deleted]]

______________________________________________
R-help@... mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

[Gabor Csardi]

Solomon, sorry for the delay. In igraph vertices are numbered from 0, so you need

keep <- 0:3

and then the function i've sent seems to work. But i can see that you also have a solution
now.

Gabor

On Sat, May 17, 2008 at 09:32:27AM -0400, Messing, Solomon O. wrote:

> Consider the following two mode-data:
>  
> edgelist:
>   actor event
> 1   Sam     a
> 2   Sam     b
> 3   Sam     c
> 4  Greg     a
> 5   Tom     b
> 6   Tom     c
> 7   Tom     d
> 8  Mary     b
> 9  Mary     d
>  
> Two-Mode Adjacency Matrix:
>      a b c d
> Sam  1 1 1 0
> Greg 1 0 0 0
> Tom  0 1 1 1
> Mary 0 1 0 1
>  
> To transform two mode to one mode data, we need a function that transforms the
> data like so:
>  
> Sam is connected to Greg (via event a)
> Sam is connected to Tom (via event b and c)
> Sam is connected to Mary (via event b)
> Tom is connected to Mary (via event b and d)
>  
> OK, now I load my data by executing the following:
> ###############################################################################
> ####
> require(igraph)
> df <- data.frame(actor = c
> ('Sam','Sam','Sam','Greg','Tom','Tom','Tom','Mary','Mary'),
>                   event =c('a','b','c','a','b','c','d','b','d') )
> g = graph.data.frame(df, directed=F)  #Coerce data to igraph object 'g'
>
> #Loading Function two.to.one:
> ##two.to.one() transforms 2-mode data to 1-mode
> two.to.one <- function(g, keep) {
>  neis <- neighborhood(g, order=2)
>  neis <- lapply(seq(neis), function(x) neis[[x]][ neis[[x]] != x-1]) ## drop
> self-loops
>  neis <- lapply(neis, function(x) x[ x %in% keep ])                  ## keep
> only these
>  neis <- lapply(seq(neis), function(x) t(cbind(x-1, neis[[x]])))     ## create
> edge lists
>  neis[-keep-1] <- NULL                                               ## these
> are not needed
>  neis <- matrix(unlist(neis), byrow=TRUE, nc=2)                      ## a
> single edge list
>  neis <- neis[ neis[,1] > neis[,2], ]                                ## count
> an edge once only
>  mode(neis) <- "character"
>  g2 <- graph.edgelist(neis, dir=FALSE)
>  V(g2)$id <- V(g2)$name          ## 'id' is used in Pajek
>  g2
> }
>  
> #Actors are the first 4 verticies, set them to be kept:
> keep = V(g)[1:4]                
> #Convert matrix with two.to.one:
> g2 = two.to.one(g, keep)
> g2
> ###############################################################################
> ####
> This yields the following output:
> > g2
> Vertices: 4
> Edges: 2
> Directed: FALSE
> Edges:
>          
> [0] 3 -- 2
> [1] 4 -- 1
>
> But, this can't be right.  Here there are only two edges where there should be
> four, and if I am inturpreting correctly, the output it is reporting that Tom
> is connected to Greg (he is not) and Sam is connected to Mary (which is true).
>  
> When I load my function, which is designed to transform a two mode edgelist
> (e.g. two columns of data) into a one-mode adjacency matrix it seems to work:
> ###############################################################################
> ####
> #load my function
> df.to.nxn <- function( x, y )
> {                                                     # x values will be the N
> x N values  
>     M <- matrix( nrow = length( unique( x ) ), ncol = length( unique( x ) ),
>           dimnames = list( unique( x ), unique( x ) ) )
>     M[ 1:length( unique( x ) ), 1:length( unique( x ) ) ] <-
> 0                    #initialize the values to 0 - this possibly could be
> removed for illustrative purposes  
>     for( i in 1:length( x ) )
> {                                                   # iterate through rows of
> data  
>             index = which( y == y[i] )                  
>                 M[ as.character( x[ index ] ), as.character( x[ index ] ) ] =
> 1      
>      }
> M                                                                                
> # return M, an N x N matrix
> }
> #Convert matrix
> g3 = df.to.nxn(df$actor, df$event)
> g4 = graph.adjacency(g3, mode = "undirected", diag = F)
> V(g4)$name = row.names(g3)
> g4
> ###############################################################################
> ####
> This yields:
> > g4
> Vertices: 4
> Edges: 4
> Directed: FALSE
> Edges:
>                
> [0] Sam  -- Greg
> [1] Sam  -- Tom
> [2] Sam  -- Mary
> [3] Tom  -- Mary
>  
> Which is what we wanted.  I have not figured out how to weight edges yet (the
> Sam and Tom edge and the Tom and Mary edge should perhaps be weighted at 2
> because 'connected twice' -- connected by two events).
>  
> -Solomon
>
> ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
> From: Gabor Csardi [mailto:csardi@...]
> Sent: Wed 5/14/2008 4:01 AM
> To: Messing, Solomon O.
> Cc: R Help list
> Subject: Re: [R] For Social Network Analysis-Graph Analysis - How to convert 2
> mode data to 1 mode data?
>
>
> Please stay on the list.
>
> On Tue, May 13, 2008 at 06:05:15PM -0400, Messing, Solomon O. wrote:
> > Gabor,
> >
> > By the way, this seems to work:
>
> I'm a bit lost. So now you're converting your data frame
> to a matrix? Why? Or you're doing the two-mode to one-mode
> conversion here? It does not seem so to me.
>
> Btw. there is a get.adjacency function in igraph to convert
> a graph to an adjacency matrix.
>
> G.
>
> >
> > df.to.nxn <- function( x, y ){
> > # x values will be the N x N values  
> >     M <- matrix( nrow = length( unique( x ) ), ncol = length( unique( x
> > ) ),
> >           dimnames = list( unique( x ), unique( x ) ) )
> >     M[ 1:length( unique( x ) ), 1:length( unique( x ) ) ] <- 0
> > # initialize the values to 0
> >     for( i in 1:length( x ) ) {
> > # iterate through rows of data  
> >             index = which( y == y[i] )                  
> >                 M[ as.character( x[ index ] ), as.character( x[ index ]
> > ) ] = 1      
> >      }
> > M
> > # return M, an N x N matrix
> > }
>
> --
> Csardi Gabor <csardi@...>    UNIL DGM
>

--
Csardi Gabor <csardi@...>    UNIL DGM

______________________________________________
R-help@... mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

[messings]

Hi Gabor,
 
Thanks for clearing that up.  There still seems to be a problem, though.  After executing:
 
###############################################################################
require(igraph)
df <- data.frame(actor = c
('Sam','Sam','Sam','Greg','Tom','Tom','Tom','Mary','Mary'),
                   event =c('a','b','c','a','b','c','d','b','d') )
g = graph.data.frame(df, directed=F)  #Coerce data to igraph object 'g'
#Loading Function two.to.one:
##two.to.one() transforms 2-mode data to 1-mode
two.to.one <- function(g, keep) {
 neis <- neighborhood(g, order=2)
 neis <- lapply(seq(neis), function(x) neis[[x]][ neis[[x]] != x-1]) ## drop
self-loops
 neis <- lapply(neis, function(x) x[ x %in% keep ])                  ## keep
only these
 neis <- lapply(seq(neis), function(x) t(cbind(x-1, neis[[x]])))     ## create
edge lists
 neis[-keep-1] <- NULL                                               ## these
are not needed
 neis <- matrix(unlist(neis), byrow=TRUE, nc=2)                      ## a
single edge list
 neis <- neis[ neis[,1] > neis[,2], ]                                ## count
an edge once only
 mode(neis) <- "character"
 g2 <- graph.edgelist(neis, dir=FALSE)
 V(g2)$id <- V(g2)$name          ## 'id' is used in Pajek
 g2
}

#Actors are the first 4 verticies, set them to be kept:
 keep = V(g)[1:4]              
#Convert matrix with two.to.one:
 g2 = two.to.one(g, keep)
V(g2)$name = V(g)$name[1:4]
g2
##############################################################33

I get the following output:

Vertices: 4
Edges: 4
Directed: FALSE
Edges:
               
[0] Sam  -- Greg
[1] Greg -- Tom
[2] Greg -- Mary
[3] Tom  -- Mary

This appears to be because the new igraph object is labeled correctly, but is not in order:

> g2 = two.to.one(g, keep)
> g2
Vertices: 4
Edges: 4
Directed: FALSE
Edges:
         
[0] 1 -- 0
[1] 0 -- 2
[2] 0 -- 3
[3] 2 -- 3

> V(g2)$name
[1] "1" "0" "2" "3"

So when I try to label the new object, the labels are assigned to the verticies incorrectly.  What is the best solution?  Could I create a new graph by ordering g2 by id?  

Thanks,

Solomon




________________________________

From: Gabor Csardi [mailto:csardi@...]
Sent: Tue 5/20/2008 3:08 AM
To: Messing, Solomon O.
Cc: R Help list
Subject: Re: [R] For Social Network Analysis-Graph Analysis - How to convert 2 mode data to 1 mode data?



Solomon, sorry for the delay. In igraph vertices are numbered from 0, so you need

keep <- 0:3

and then the function i've sent seems to work. But i can see that you also have a solution
now.

Gabor

On Sat, May 17, 2008 at 09:32:27AM -0400, Messing, Solomon O. wrote:

> Consider the following two mode-data:
>
> edgelist:
>   actor event
> 1   Sam     a
> 2   Sam     b
> 3   Sam     c
> 4  Greg     a
> 5   Tom     b
> 6   Tom     c
> 7   Tom     d
> 8  Mary     b
> 9  Mary     d
>
> Two-Mode Adjacency Matrix:
>      a b c d
> Sam  1 1 1 0
> Greg 1 0 0 0
> Tom  0 1 1 1
> Mary 0 1 0 1
>
> To transform two mode to one mode data, we need a function that transforms the
> data like so:
>
> Sam is connected to Greg (via event a)
> Sam is connected to Tom (via event b and c)
> Sam is connected to Mary (via event b)
> Tom is connected to Mary (via event b and d)
>
> OK, now I load my data by executing the following:
> ###############################################################################
> ####
> require(igraph)
> df <- data.frame(actor = c
> ('Sam','Sam','Sam','Greg','Tom','Tom','Tom','Mary','Mary'),
>                   event =c('a','b','c','a','b','c','d','b','d') )
> g = graph.data.frame(df, directed=F)  #Coerce data to igraph object 'g'
>
> #Loading Function two.to.one:
> ##two.to.one() transforms 2-mode data to 1-mode
> two.to.one <- function(g, keep) {
>  neis <- neighborhood(g, order=2)
>  neis <- lapply(seq(neis), function(x) neis[[x]][ neis[[x]] != x-1]) ## drop
> self-loops
>  neis <- lapply(neis, function(x) x[ x %in% keep ])                  ## keep
> only these
>  neis <- lapply(seq(neis), function(x) t(cbind(x-1, neis[[x]])))     ## create
> edge lists
>  neis[-keep-1] <- NULL                                               ## these
> are not needed
>  neis <- matrix(unlist(neis), byrow=TRUE, nc=2)                      ## a
> single edge list
>  neis <- neis[ neis[,1] > neis[,2], ]                                ## count
> an edge once only
>  mode(neis) <- "character"
>  g2 <- graph.edgelist(neis, dir=FALSE)
>  V(g2)$id <- V(g2)$name          ## 'id' is used in Pajek
>  g2
> }
>
> #Actors are the first 4 verticies, set them to be kept:
> keep = V(g)[1:4]              
> #Convert matrix with two.to.one:
> g2 = two.to.one(g, keep)
> g2
> ###############################################################################
> ####
> This yields the following output:
> > g2
> Vertices: 4
> Edges: 2
> Directed: FALSE
> Edges:
>        
> [0] 3 -- 2
> [1] 4 -- 1
>
> But, this can't be right.  Here there are only two edges where there should be
> four, and if I am inturpreting correctly, the output it is reporting that Tom
> is connected to Greg (he is not) and Sam is connected to Mary (which is true).
>
> When I load my function, which is designed to transform a two mode edgelist
> (e.g. two columns of data) into a one-mode adjacency matrix it seems to work:
> ###############################################################################
> ####
> #load my function
> df.to.nxn <- function( x, y )
> {                                                     # x values will be the N
> x N values
>     M <- matrix( nrow = length( unique( x ) ), ncol = length( unique( x ) ),
>           dimnames = list( unique( x ), unique( x ) ) )
>     M[ 1:length( unique( x ) ), 1:length( unique( x ) ) ] <-
> 0                    #initialize the values to 0 - this possibly could be
> removed for illustrative purposes  
>     for( i in 1:length( x ) )
> {                                                   # iterate through rows of
> data  
>             index = which( y == y[i] )                  
>                 M[ as.character( x[ index ] ), as.character( x[ index ] ) ] =
> 1      
>      }
> M                                                                              
> # return M, an N x N matrix
> }
> #Convert matrix
> g3 = df.to.nxn(df$actor, df$event)
> g4 = graph.adjacency(g3, mode = "undirected", diag = F)
> V(g4)$name = row.names(g3)
> g4
> ###############################################################################
> ####
> This yields:
> > g4
> Vertices: 4
> Edges: 4
> Directed: FALSE
> Edges:
>              
> [0] Sam  -- Greg
> [1] Sam  -- Tom
> [2] Sam  -- Mary
> [3] Tom  -- Mary
>
> Which is what we wanted.  I have not figured out how to weight edges yet (the
> Sam and Tom edge and the Tom and Mary edge should perhaps be weighted at 2
> because 'connected twice' -- connected by two events).
>
> -Solomon
>
> ???????????????????????????????????????????????????????????????????????????????
> From: Gabor Csardi [mailto:csardi@...]
> Sent: Wed 5/14/2008 4:01 AM
> To: Messing, Solomon O.
> Cc: R Help list
> Subject: Re: [R] For Social Network Analysis-Graph Analysis - How to convert 2
> mode data to 1 mode data?
>
>
> Please stay on the list.
>
> On Tue, May 13, 2008 at 06:05:15PM -0400, Messing, Solomon O. wrote:
> > Gabor,
> >
> > By the way, this seems to work:
>
> I'm a bit lost. So now you're converting your data frame
> to a matrix? Why? Or you're doing the two-mode to one-mode
> conversion here? It does not seem so to me.
>
> Btw. there is a get.adjacency function in igraph to convert
> a graph to an adjacency matrix.
>
> G.
>
> >
> > df.to.nxn <- function( x, y ){
> > # x values will be the N x N values
> >     M <- matrix( nrow = length( unique( x ) ), ncol = length( unique( x
> > ) ),
> >           dimnames = list( unique( x ), unique( x ) ) )
> >     M[ 1:length( unique( x ) ), 1:length( unique( x ) ) ] <- 0
> > # initialize the values to 0
> >     for( i in 1:length( x ) ) {
> > # iterate through rows of data  
> >             index = which( y == y[i] )                  
> >                 M[ as.character( x[ index ] ), as.character( x[ index ]
> > ) ] = 1      
> >      }
> > M
> > # return M, an N x N matrix
> > }
>
> --
> Csardi Gabor <csardi@...>    UNIL DGM
>

--
Csardi Gabor <csardi@...>    UNIL DGM



        [[alternative HTML version deleted]]

______________________________________________
R-help@... mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.