Network analysis (R).Rmd

---
title: Network analysis (R)
output: html_document
editor_options: 
  markdown: 
    wrap: 72
---

# Network analysis

Network analysis focuses on seeing the world via nodes and ties between
nodes. It can be used for various purposes, ranging from [textual
data](https://sicss.io/2020/materials/day3-text-analysis/text-networks/rmarkdown/Text_Networks.html)
to analysis of [sexual relationships in a high
school](https://people.duke.edu/~jmoody77/chains.pdf).

Now we learn how to \* read data in as a network \* visualise network \*
compute various network characteristics

For this, we use the package [igraph](https://igraph.org/r/).

[Narrated code walkthrough](https://www.youtube.com/watch?v=lXC2egj8lmY)

```{r}
library(igraph)
```

```{r}
data <- read.csv('data/org_x_collaboration.csv', header = F )
matrix <- as.matrix( data )

network <- graph_from_adjacency_matrix( matrix, 'directed' )
```

## Computing network characteristics

There are various measurements which can be executed both at the level
of individual nodes as well as at the level of the whole network.

```{r}
degree( network )
summary( degree( network ) )
hist( degree( network ) )
```

## Exercises

-   Instead of this ugly histogram, use `ggplot` to draw a nice
    histogram.
-   Instead of degree use
    [betweness](https://igraph.org/r/doc/betweenness.html) metrics.
-   Do degree and betweness correlate?
-   What is the diameter of the network?

## Plotting networks

Often network analysis is done through drawing the network. Here we
learn how to twist the drawing by changing node attributes, tie
attributes as well as the [layout of the
graph](https://igraph.org/r/doc/layout_.html). See
[documentation](https://igraph.org/r/doc/plot.common.html) for various
alternatives you can use.

```{r}
plot( network )
```

```{r}
V(network)$color <- 'white'
V(network)$size <- degree(network)

E(network)$color <- 'green'

l = layout_in_circle( network )

plot( network, layout = l )
```

## Exercises

-   Draw a network which you would use to describe the outcomes to the
    organisation.
-   Examine different layout algorithms and see how different they are.

## Community detection

Beyond visualizations, one can try to find communities (or clusters) of
more dencly connected nodes from the materials. There are various
community detection algorithms out there [see for
example](https://www.rdocumentation.org/packages/igraph/versions/1.3.5/topics/membership),
and they do produce somewhat different outcomes.

```{r}
communities <- cluster_optimal( network )
V(network)$community <- membership( communities )

print( V(network)$community )

V(network)$color <- V(network)$community
V(network)$size <- degree(network)

E(network)$color <- 'green'

l = layout_in_circle( network )

plot( network, layout = l )
```

## Exercises

-   Try out different clustering approaches: what kinds of insights they
    provide to the data?