network
/sna
edition)For years now, authors and analysts have worked on financial data
using ad-hoc tools or programming languages other than
R
. So, the package FinNet
was born to provide
all R
users with the ability to study financial networks
with a set of tool especially designed to this purpose. Specifically,
FinNet
offers both brand new tools and an interface to the
almost limitless capabilities of igraph
and
network
.
This vignette illustrates how to:
yahoofinancer
;After having identified the firms of interest, the package can fetch
all information on them as long as yahoofinancer
is
available. Otherwise, built-in data can be used:
There are many function in the FF
function
family to rapidly build an adjacency matrix. In this step,
FF.norm.ownership()
will construct a normalised-valued
matrix of common ownership
A graph can be obtained easily using FF.net()
, which
include two preset aesthetics: ‘simple’ and ‘nice’
Some checks using the S3 methods implemented for
financial_matrix
objects and the extension of some
igraph
functions allow to verify the correctness of the
graph:
# The order of the graph equals the number of rows in the FF matrix
network.size(net) == nrow(FF)
#> [1] TRUE
# The names of its vertex match the row names of the FF matrix
vertex.names(net) == rownames(FF)
#> [1] TRUE TRUE TRUE
# Plot it
plot_network(net, displaylabels = TRUE, vertex.cex = 3, edge.lwd = 6)
The ‘nice’ defaults are more indicated for a visual inspection of the network. net
# Load dataset
data('firms_BKB')
# Identify common-ownership relations in a firm-firm matrix
FF <- FF(firms_BKB, who = 'own',
ties = 'naive', Matrix = TRUE)
# Create a nice-looking graph
net <- FF.net(FF, aesthetic = 'nice')
# Plot it
plot_network(net, displaylabels = FALSE, edge.col = 'darkgray',
vertex.cex = 3, edge.lwd = 1)