gvdr/unravel

Plots Meaningful Representations of Complex Networks

unravel

Hairballs, hairballs everywhere. Most complex networks are plot in an (arguably) awe-inspiring, questionably informative way, that we should call “a ball of yarn” (Guðmundsdóttir B. Unravel, in Homogenic. El Cortijo, Málaga, Spain. 1997 Sept 22). A lot of color (that don’t mean anything), a lot of node-links-overlap mess (that is hard to read into), and a lot of force-directed graph layout algorithms that nobody could ever explain me why should they be used a part that “hey, they are cool!”.

The result of this is that the graphic parameters of a network representation are almost never meaningful. You don’t know what you are looking at.

The same is true also for food webs, that is the domain of my research. I know food webs (well, a little at least), I’ve been working on food webs for the last 8 years. Yet, if you show me a food web plot I would not “see” anything.

Here comes unravel!

The radical philosophy of unravel dictates that:

all graphic elements (e.g., size, colour, placement) should mean something

This means, for example, that the vertical and horizontal position of a node in the plot must correspond to something. This, requires domain knowledge (and that’s the reason why I started with food webs, but I hope somebody else will help me with other kind of networks).

This also means that if something does not mean anything, it must not be used. So, if you can assign a vertical / horizontal position to your nodes based on something you know about the nodes, think twice: do you reallY want to plot the network as a hairball? What information will the reader get from your plot? Isn’t there a better way of plotting it?

Installation

unravel is not yet on CRAN. You can install it from GitHub with ease:

remotes::install_github("gvdr/unravel")

A word about GitHub. GitHub has systematically banned people coming from or living in certain countries: that sucks, and I’m looking for a better place where to transfer the package.

Example

The main high level plotting function produces a specialised, domain aware, and opinionated version of a ggraph plot. That means that (1) it makes decision and (2) if you don’t make decision, unravel will make decision for you.

library(unravel)
library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
## basic example code

data("TuesdayLakes_1984")

TuesdayLakes_1984 %>%
  plot_foodweb(dodge = 0.08)

Additional graphical elements can be addedd using the ggraph (and ggplot2) API.

library(viridis)
#> Loading required package: viridisLite
library(ggraph)
#> Loading required package: ggplot2


TuesdayLakes_1984 %>%
  plot_foodweb(dodge = 0.08) +
  ggtitle("Tuesday Lake 1984") +
  scale_colour_viridis()

The underlying, more flexible, function is ggfoodweb(). It requires both a foodweb and a layout (that is, a matrix with two columns, specifying the x and y position of the specis in the plot). ggfoodweb() does not compute the layer for you, nor provides colour and size for the nodes. You can use tl_oi_layer() to compute a layout, and fw_node_{size,colour}() to define size and colour.

In the following example, we change size and colour with two different centralities, and place the species on the x axis following the order in which they are written in the network (this is usually a very BAD choice).

default_layer <- tl_oi_layer(TuesdayLakes_1984, dodge = 0.04)

# we change the first column (horizontal position) to the row number of species
default_layer[,1] <- 1:nrow(default_layer) 


TuesdayLakes_1984 %>%
  fw_node_colour(tidygraph::centrality_eigen) %>%
  fw_node_size(tidygraph::centrality_authority) %>%
  ggfoodweb(default_layer) +
  # with ggfoodweb() we need to define the axis and guides labels by hand
  scale_x_continuous(name = "Species Index") +
  scale_y_continuous(name = "Trophic Level") +
  guides(size = guide_legend(title="Authority"),
         colour = guide_colourbar(title="Eigen Centr")) +
  ggtitle("Tuesday Lake 1984") +
  scale_colour_viridis()

Code of Conduct and Contribution

Please note that the ‘unravel’ project is released with a Contributor Code of Conduct. By contributing to this project, you agree to abide by its terms.