Notes from a from a plant biologist - Sankey diagram example data and code

I first learned about a Sankey graph from Tufte’s data visualization book (Tufte 1983). It is a heartbreaking image of Napoleon’s army’s attempt to invade Moscow, but instead of victory, thousands died through a mix of war and unhygienic and freezing weather as the marched to Moscow and back to France. My purposes for this post are much more mundane, quite unimportant relatively - to look at weightings of a project with criteria.

How can data be used to generate a Sankey graph? You need project and criteria node names arranged in terms of weights and the target and sources for links Table 2. This way nodes and links are sized according the assigned weights Figure 1 - an example of its application is in my other post for our weed horizon scan.

Charles Joseph Minard’s visualization of Napoleon’s Russian campaign of 1812

The package is a network package networkD3Allaire et al. (2017).

#Load packages
library(tidyverse)
library(networkD3)
library(kableExtra)

The data can be read in from an external file, you need the nodes to be listed and numbered from 0 and on up to the number of nodes in one file, the names should match the criteria and project titles Table 1.

nodes<-
  structure(
  list(
    source = 0:7, #node numbering starts at zero
    target_name = c(
      "1 project",
      "2 project",
      "3 project",
      "4 project",
      "5 project",
      "Proportion of work that is plant science",
      "Stakeholders will benefit",
      "Is not well addressed by others"
    )
  ),
  row.names = c(NA,-8L),
  class = "data.frame"
)

kable(nodes %>% arrange(source), align="l", format = "pipe", padding = 2)

Table 1: The nodes need to be numbered from zero up.
source	target_name
0	1 project
1	2 project
2	3 project
3	4 project
4	5 project
5	Proportion of work that is plant science
6	Stakeholders will benefit
7	Is not well addressed by others

Generate some random weights for this exercise, I am using 3 criteria, for the 5 projects so I need 15 weights.

weights<-sample(1:10, 15, replace=T)

Then generate the table with project names and weights. The link colors are determined by the names on the nodes and it can have problems recognizing names as unique if the first word is the same and there are spaces. I used Project 1, Project 2, etc and it didn’t recognize the nodes as distinct until I named them 1 Project, 2 Project, etc.

target_source <-
  tibble(
    source_name = c(rep(
      c(
        "Proportion of work that is plant science",
        "Stakeholders will benefit",
        "Is not well addressed by others"
      ),
      5
    )),
    target_name = c(rep(
      c("1 project",
        "2 project",
        "3 project",
        "4 project",
        "5 project"),
      3
    )),
    weight=as.numeric(c(weights)),
    target = c(rep(c(0, 1, 2, 3, 4), 3)),
    source = c(rep(c(5, 6, 7), 5))
    
    ) 

kable(target_source %>% arrange(target_name), align="c", format = "pipe", padding = 2)

Table 2: Data arrangement includes projects, criteria and weights.
source_name	target_name	weight	target	source
Proportion of work that is plant science	1 project	3	0	5
Is not well addressed by others	1 project	9	0	7
Stakeholders will benefit	1 project	8	0	6
Stakeholders will benefit	2 project	7	1	6
Proportion of work that is plant science	2 project	3	1	5
Is not well addressed by others	2 project	10	1	7
Is not well addressed by others	3 project	6	2	7
Stakeholders will benefit	3 project	10	2	6
Proportion of work that is plant science	3 project	9	2	5
Proportion of work that is plant science	4 project	7	3	5
Is not well addressed by others	4 project	4	3	7
Stakeholders will benefit	4 project	7	3	6
Stakeholders will benefit	5 project	5	4	6
Proportion of work that is plant science	5 project	2	4	5
Is not well addressed by others	5 project	7	4	7

Then generate the Figure 1. Just note it is outputting and html widget, looks a lot better on a desktop screen.

sankey3<-networkD3::sankeyNetwork(Links = target_source, 
                                  Nodes = nodes, 
                                  Source = 'target', 
                                  Target = 'source', 
                                  Value = 'weight', 
                                  NodeID = 'target_name',
                                  units = 'weight',
                                  width = 600,
                                  height = 800,
                                  fontSize = 18,
                                  LinkGroup="target_name", 
                                  fontFamily = "Georgia")

sankey3

Figure 1: Project weights under three criteria using a Sankey diagram, from R code.

References

Allaire JJ, Gandrud C, Russell K, Yetman C 2017. networkD3: D3 JavaScript network graphs from r.

Tufte ER 1983. The visual display of quantitative information. Graphics Press, Chesire, CT.