图形边缘宽度与颜色正负相关

Question

I´m trying to plot an i graph using taxonomic information. The original correlation matrix is attached as a dput output.

The original dput file correlation matrixe is found here

My code looks like this. Assuming cor.matrix is the correlation matrix (see link to get it).

set.seed(123)

  t = which(abs(cor.matrix) > 0.6 & lower.tri(cor.matrix),arr.ind=TRUE)
  t.graph=graph.data.frame(t,directed=F)
  E(t.graph)$color =ifelse(cor.matrix[t] > 0.6,'magenta','green')

  t.names <- colnames(cor.matrix)[as.numeric(V(t.graph)$name)]
  minC <- rep(-Inf, vcount(t.graph))
  maxC <- rep(Inf, vcount(t.graph))
  minC[1] <- maxC[1] <- 0
  l <- layout_with_fr(t.graph, minx=minC, maxx=maxC,
                       miny=minC, maxy=maxC)      
  plot(t.graph, layout=l, 
       rescale=T,
       asp=0,
       edge.arrow.size=0.5, 
       vertex.label.cex=0.8, 
       vertex.label.family="Helvetica",
       vertex.label.font=2,
       vertex.label=t.names,
       vertex.shape="circle", 
       vertex.size=3, 
       vertex.color="deepskyblue2",
       vertex.label.color="black", 
       edge.width=0.5)

I would like to accomplish three different things:

1 - Have one color for positive and negative correlation.

2 - Change edge width, the higher the correlation (positive or negative) the thicker the edge.

3 - Have a graph easier to vew, it´s hard to view and identify nodes

Thanks so much.

Answer 1

For points 1 and 2 you just need to pass a vector to the appropriate plotting argument. For edge color use edge.color and for edge width use edge.width you can also set an edge attribute in the graph object and igraph will automatically use that for plotting. The igraph manual specifies a lot of this: http://igraph.org/r/doc/plot.common.html

As for point 3. The only way to make the graph more readable is to make plot it to a bigger canvas (ie increase the resolution) or remove some nodes. Graphs become notoriously hard to read as the number of nodes increases and there is nothing to do about it.

library(igraph)
set.seed(123)

cor.matrix <- matrix(runif(100, -1, 1), nrow=10)

t = which(abs(cor.matrix) > 0.6 & lower.tri(cor.matrix),arr.ind=TRUE)
t <- cbind(t, cor.matrix[which(abs(cor.matrix) > 0.6 & lower.tri(cor.matrix),arr.ind=TRUE)]) ##this adds the correlation to the graph as an edge attribute "V3"
t.graph=graph.data.frame(t,directed=F)
E(t.graph)$color <- ifelse(E(t.graph)$V3 > 0,'magenta','green') #You had this as "V3 > 0.6" which I guess works but it is more readable as 0. that way if you decide to lower the correlation threshold you do not have to change this line too.

#t.names <- colnames(cor.matrix)[as.numeric(V(t.graph)$name)]
minC <- rep(-Inf, vcount(t.graph))
maxC <- rep(Inf, vcount(t.graph))
minC[1] <- maxC[1] <- 0
l <- layout_with_fr(t.graph, minx=minC, maxx=maxC,
                    miny=minC, maxy=maxC)      
plot(t.graph, layout=l, 
     rescale=T,
     asp=0,
     edge.arrow.size=0.5, 
     vertex.label.cex=0.8, 
     vertex.label.family="Helvetica",
     vertex.label.font=2,
     #vertex.label=t.names,
     vertex.shape="circle", 
     vertex.size=3, 
     vertex.color="deepskyblue2",
     vertex.label.color="black", 
     #edge.color=E(t.graph)$color, ##do not need this since E(t.graph)$color is already defined.
     edge.width=as.integer(cut(abs(E(t.graph)$V3), breaks = 5)))