CS&SS 569 Visualizing Data and Models

.title[
# CS&SS 569 Visualizing Data and Models
]
.subtitle[
## Lab 7: Visualizing relational data
]
.author[
### Brian Leung
]
.institute[
### Department of Political Science, UW
]
.date[
### 2023-02-17
]

---

## Introduction

- Cover different ways to visualize network or relational data

- Node-edge diagram (classic network visualization) with **`tidygraph`** and **`ggraph`**

- Heat map (plus cluster analysis) with **`ggplot2`** and **`cluster`**

- Increasingly popular ways

- Chord diagram with **`circlize`**

- Sankey diagrams with **`ggsankey`** (or alluvial diagram with **`ggalluvial`**) 
  
---

## Network visualization

- Map of sciences (Bollen et al. 2009) based on clickstream data on journal websites

---

## Challenges

- Network data create many challenges for visualization

- Cursed by high dimensionality

- Network diagrams usually result in hairballs or spaghetti balls. . .

- One takeaway from this lab is to take advantage of alternative visualization methods when possible

- Not just network data, but more generically, *relational* data

- Any data whose unit of observation is dyadic

- E.g., migration flow data, or import/export data, between a pair of countries

---
  
## Examples in class

- Florentine families and the rise of Medici: node-edge diagram

---

## Examples in class

- Migration flow data: heat map with cluster analysis

---

## Examples in class

- Migration flow data: chord diagram

---

## Examples in class

- Migration flow data: sankey diagram

---

## Prerequisite

```r
# download packages 
packages <- c("tidyverse", "tidygraph", "ggraph", "reshape2", "cluster", "circlize")
not_installed <- setdiff(packages, rownames(installed.packages()))
if (length(not_installed)) install.packages(not_installed)

# install developmental version of ggsankey
devtools::install_github("davidsjoberg/ggsankey")

# load packages
library(tidyverse)
library(tidygraph)
library(ggraph)
library(reshape2)
library(cluster)
library(circlize)
library(ggsankey)
```

---

```r
# load data
*medici <- read.table("data/medici.txt")
```
]
 
.panel2-medici1-auto[

]

---
count: false
 
## Rise of Medici: Adjacency Matrix
.panel1-medici1-auto[

```r
# load data
medici <- read.table("data/medici.txt")

# examine the data
*head(medici)
```
]
 
.panel2-medici1-auto[

```
           Acciaiuoli Albizzi Barbadori Bischeri Castellani Ginori Guadagni
Acciaiuoli          0       0         0        0          0      0        0
Albizzi             0       0         0        0          0      1        1
Barbadori           0       0         0        0          1      0        0
Bischeri            0       0         0        0          0      0        1
Castellani          0       0         1        0          0      0        0
Ginori              0       1         0        0          0      0        0
           Lamberteschi Medici Pazzi Peruzzi Pucci Ridolfi Salviati Strozzi
Acciaiuoli            0      1     0       0     0       0        0       0
Albizzi               0      1     0       0     0       0        0       0
Barbadori             0      1     0       0     0       0        0       0
Bischeri              0      0     0       1     0       0        0       1
Castellani            0      0     0       1     0       0        0       1
Ginori                0      0     0       0     0       0        0       0
           Tornabuoni
Acciaiuoli          0
Albizzi             0
Barbadori           0
Bischeri            0
Castellani          0
Ginori              0
```
]

---

count: false
 
## Rise of Medici: tidygraph package for manipulation
.panel1-medici2-auto[

```r
# coerce into a tbl_graph using tidygraph
*medici_graph <- as_tbl_graph(medici, directed = FALSE)
```
]
 
.panel2-medici2-auto[

]

---
count: false
 
## Rise of Medici: tidygraph package for manipulation
.panel1-medici2-auto[

```r
# coerce into a tbl_graph using tidygraph
medici_graph <- as_tbl_graph(medici, directed = FALSE)

# examine the tbl_graph
*medici_graph
```
]
 
.panel2-medici2-auto[

```
# A tbl_graph: 16 nodes and 58 edges
#
# An undirected multigraph with 2 components
#
# Node Data: 16 × 1 (active)
  name      
  <chr>     
1 Acciaiuoli
2 Albizzi   
3 Barbadori 
4 Bischeri  
5 Castellani
6 Ginori    
# … with 10 more rows
#
# Edge Data: 58 × 3
   from    to weight
  <int> <int>  <dbl>
1     1     1      1
2     1     2      1
3     1     3      1
# … with 55 more rows
```
]

---

count: false
 
## Rise of Medici: Basic visualization w/ ggraph
.panel1-medici3-auto[

```r
*ggraph(medici_graph)
```
]
 
.panel2-medici3-auto[
![](lab7_slides_files/figure-html/medici3_auto_01_output-1.png)
]

---
count: false
 
## Rise of Medici: Basic visualization w/ ggraph
.panel1-medici3-auto[

```r
ggraph(medici_graph) +
* geom_node_point()
```
]
 
.panel2-medici3-auto[
![](lab7_slides_files/figure-html/medici3_auto_02_output-1.png)
]

---
count: false
 
## Rise of Medici: Basic visualization w/ ggraph
.panel1-medici3-auto[

```r
ggraph(medici_graph) +
  geom_node_point() +
* geom_edge_link()
```
]
 
.panel2-medici3-auto[
![](lab7_slides_files/figure-html/medici3_auto_03_output-1.png)
]

---
count: false
 
## Rise of Medici: Basic visualization w/ ggraph
.panel1-medici3-auto[

```r
ggraph(medici_graph) +
  geom_node_point() +
  geom_edge_link() +
* geom_node_text(aes(label = name), repel = TRUE)
```
]
 
.panel2-medici3-auto[
![](lab7_slides_files/figure-html/medici3_auto_04_output-1.png)
]

---
count: false
 
## Rise of Medici: Basic visualization w/ ggraph
.panel1-medici3-auto[

```r
ggraph(medici_graph) +
  geom_node_point() +
  geom_edge_link() +
  geom_node_text(aes(label = name), repel = TRUE) +
* theme_graph()
```
]
 
.panel2-medici3-auto[
![](lab7_slides_files/figure-html/medici3_auto_05_output-1.png)
]

---

count: false
 
## Rise of Medici: tidygraph package for network computation
.panel1-medici4-auto[

```r
*medici_graph
```
]
 
.panel2-medici4-auto[

---
count: false
 
## Rise of Medici: tidygraph package for network computation
.panel1-medici4-auto[

```r
medici_graph %>%
* mutate(
*   degree = centrality_degree(),  # degree centrality
*   community = group_leading_eigen()  # community-detection algorithm
* )
```
]
 
.panel2-medici4-auto[

```
# A tbl_graph: 16 nodes and 58 edges
#
# An undirected multigraph with 2 components
#
# Node Data: 16 × 3 (active)
  name       degree community
  <chr>       <dbl>     <int>
1 Acciaiuoli      7         1
2 Albizzi         9         1
3 Barbadori       9         1
4 Bischeri        9         2
5 Castellani      7         2
6 Ginori          4         3
# … with 10 more rows
#
# Edge Data: 58 × 3
   from    to weight
  <int> <int>  <dbl>
1     1     1      1
2     1     2      1
3     1     3      1
# … with 55 more rows
```
]

---
count: false
 
## Rise of Medici: tidygraph package for network computation
.panel1-medici4-auto[

```r
medici_graph %>%
  mutate(
    degree = centrality_degree(),  # degree centrality
    community = group_leading_eigen()  # community-detection algorithm
  ) ->
* medici_graph
```
]
 
.panel2-medici4-auto[

]

---

count: false
 
## Rise of Medici: Basic visualization w/ ggraph
.panel1-medici5-non_seq[

```r
ggraph(
  medici_graph,
  )
```
]
 
.panel2-medici5-non_seq[
![](lab7_slides_files/figure-html/medici5_non_seq_01_output-1.png)
]