| plotSoilRelationGraph {sharpshootR} | R Documentation |
Plot a component relation graph based on an adjacency or similarity matrix.
plotSoilRelationGraph(m, s='', plot.style='network', spanning.tree=NULL, del.edges=NULL, vertex.scaling.factor=2, edge.scaling.factor=1, edge.transparency=1, edge.col=grey(0.5), edge.highlight.col='royalblue', g.layout=layout.fruchterman.reingold, ...)
m |
adjacency matrix |
s |
central component; an empty character string is interpreted as no central component |
plot.style |
plot style ('network', or 'dendrogram'), or 'none' for no graphical output |
spanning.tree |
plot the minimim or maximum spaning tree ('min', 'max'), or, max spanning tree plus edges with weight greater than the n-th quantile specified in 'spanning.tree'. See details and examples. |
del.edges |
optionally delete edges with weights less than the specified quantile (0-1) |
vertex.scaling.factor |
scaling factor applied to vertex size |
edge.scaling.factor |
optional scaling factor applied to edge width |
edge.transparency |
optional transparency setting for edges (0-1) |
edge.col |
edge color, applied to all edges |
edge.highlight.col |
edge color applied to all edges connecting to component named in |
g.layout |
an igraph layout function, defaults to |
... |
further arguments passed to plotting function |
Vertex size is based on a normalized index of connectivity: size = sqrt(degree(g)/max(degree(g))) * scaling.factor. Edge width can be optionally scaled by edge weight by specifying an edge.scaling.factor value. The maximum spanning tree represents a sub-graph where the sum of edge weights are maximized. The mimimum spanning tree represents a sub-graph where the sum of edge weights are minimized. The maximum spanning tree is likely a more useful simplification of the full graph, in which only the strongest relationships (e.g. most common co-occurrences) are preserved.
The maximum spanning tree + edges with weights > n-th quantile is an experimental hybrid. The 'backbone' of the graph is created by the maximum spanning tree, and augmented by 'strong' auxillary edges– defined by a value between 0 and 1.
an igraph 'graph' object is invisibly returned
This function is a work in progress, ideas welcome.
D.E. Beaudette
# load sample data set
data(amador)
# create weighted adjacency matrix (see ?component.adj.matrix for details)
m <- component.adj.matrix(amador)
# plot network diagram, with Amador soil highlighted
plotSoilRelationGraph(m, s='amador')
# dendrogram representation
plotSoilRelationGraph(m, s='amador', plot.style='dendrogram')
# compare methods
m.o <- component.adj.matrix(amador, method='occurrence')
par(mfcol=c(1,2))
plotSoilRelationGraph(m, s='amador', plot.style='dendrogram')
title('community matrix')
plotSoilRelationGraph(m.o, s='amador', plot.style='dendrogram')
title('occurence')
# investigate max spanning tree
plotSoilRelationGraph(m, spanning.tree='max')
# investigate max spanning tree + edges with weights > 75-th pctile
plotSoilRelationGraph(m, spanning.tree=0.75)
## Not run:
##D # get similar data from soilweb, for the Pardee series
##D s <- 'pardee'
##D u <- url(URLencode(paste(
##D 'http://casoilresource.lawr.ucdavis.edu/soil_web/reflector_api/soils.php?',
##D 'what=soil_series_component_query&q_string=', s, sep='')))
##D
##D # fetch data
##D d <- read.table(u, sep='|', header=TRUE, stringsAsFactors=FALSE)
##D
##D # normalize component names
##D d$compname <- tolower(d$compname)
##D
##D # keep only major components
##D d <- subset(d, subset=compkind == 'Series')
##D
##D # build adj. matrix and plot
##D m <- component.adj.matrix(d)
##D plotSoilRelationGraph(m, s=s, plot.style='dendrogram')
##D
##D # alter plotting style, see ?plot.phylo
##D plotSoilRelationGraph(m, s=s, plot.style='dendrogram', type='fan')
##D plotSoilRelationGraph(m, s=s, plot.style='dendrogram', type='unrooted', use.edge.length=FALSE)
## End(Not run)