Vizards

Today, we are working in Markdown so you can see your work as you proceed. We will be using igraph exclusively. But all the lessons that you learn will work for any network or network visualization software.

In the end, you just want to be effective.

So, start by initiating loading the igraph library and loading the data for today.

To create a code chunk:




library(igraph)
## 
## Attaching package: 'igraph'
## The following objects are masked from 'package:stats':
## 
##     decompose, spectrum
## The following object is masked from 'package:base':
## 
##     union
# Load networks
load("data/allege.rda")
load("data/negative.rda")
load("data/positive.rda")
load("data/secrets.rda")
load("data/sentiment.rda")

# allege <- graph_from_adjacency_matrix(as.matrix(allege))

Note to self: Look iup “markdown cheat sheets” for R Studio to learn more.

At this point, you should have five networks in R’s memory: allege, negative, positive, secrets, and sentiment.

Now, you are ready to start.




Things to consider when visualizing networks

Visualization is seldom adequate for analysis

 Use visualization to illustrate analysis or develop your initial opinions.

Here is a first look at the two networks. To keep it interesting, let’s put them together in one row (and two columns).

We should also set some options now so that we don’t have to keep adding arguments to each plot.

In this case, the code chunk below is telling igraph to use the Kamada Kawaii algorithm for each visualization it draws. It will also decrease the size of the arrowheads to 0.2 (the default is 1).

igraph_options(plot.layout=layout_with_kk, 
               edge.arrow.size=0.2,
               vertex.label.cex=0.5)

#global options

Now, plot the networks.

par(mfrow=c(2,2), mar=c(0.5,0,2,0))
# make a two row, two column table of plots. margins (bottom, left, top, right) (default is 1 1 2 1)
plot(positive, main="Positive Sentiment")
plot(negative, main="Negative Sentiment")
plot(allege, main="Made Allegations About")
plot(secrets, main="Knows a Secret About")

par(mfrow=c(1,1))
#make the plot window plot just one graph at a time

The layout is ultimately arbitrary

This doesn’t mean that you can be arbitrary about which layout to select. Rather, the relative position of the nodes should not be taken as a demonstration that two nodes are necessarily similar or different. You are looking at what happens when we try to fit a multidimensional object into two dimensions.


#### There is no one “best” plotting solution.

Try out a range of options for plotting the networks. Each plotting algorithm is meant to work with a different type of network situation. For a list of the potential graphing algorithms in igraph, try:

?layout_

…or you can also try out some other options, such as manipulating the plot yourself with tkplot() or using 3-D with rglplot(). But those options are fairly rudimentary compared with some of the newer options that you have these days.

For more on that, check out visNetwork(). That will start you down a rabbit hole.


In this case, we are going to learn how to take and reuse coordinates from one plot in others.

First, take the x and y coordinates of each node using this code:

coords <- layout_with_kk(positive)

Now, let’s plot that again…

par(mfrow=c(2,2), mar=c(0.5,0,2,0))
# make a two row, two column table of plots. margins (bottom, left, top, right) (default is 1 1 2 1)
plot(positive, main="Positive Sentiment", layout=coords)
plot(negative, main="Negative Sentiment", layout=coords)
plot(allege, main="Made Allegations About", layout=coords)
plot(secrets, main="Knows a Secret About", layout=coords)

par(mfrow=c(1,1))
#make the plot window plot just one graph at a time

Now, if you like that, then you can add the coordinates to each network. We add the coordinates as an x and y attribute of the vertices. coors [rows, columns]

V(positive)$x <- coords[ ,1]
V(positive)$y <- coords[ ,2]
V(negative)$x <- coords[ ,1]
V(negative)$y <- coords[ ,2]
V(allege)$x <- coords[ ,1]
V(allege)$y <- coords[ ,2]
V(secrets)$x <- coords[ ,1]
V(secrets)$y <- coords[ ,2]

Next, remove isolates and name the new networks something unique so you don’t overwrite the original networks. They will plot according to their coordinates, unless you tell them to do something else.

Your Ultimate Priority

The priorities are always communication and clarity.
A simpler approach is generally more powerful.
Depict no more than 2 or 3 properties at a time.

Adding attributes

Attributes can add a lot of information to a visualization.

atts <- read.csv("data/character_attributes.csv", header=TRUE)

Before you try to add attributes, make sure that the nodes in the network appear in the same order as the nodes in the attribute sheet.

cbind(V(positive)$name, atts[ ,1])#bind the two into columns so we can compare them

Then, add the attributes to each network.

V(positive)$family <- atts[ ,2]
V(positive)$employed <- atts[ ,3]
V(positive)$sex <- atts[ ,4]

V(negative)$family <- atts[ ,2]
V(negative)$employed <- atts[ ,3]
V(negative)$sex <- atts[ ,4]

V(allege)$family <- atts[ ,2]
V(allege)$employed <- atts[ ,3]
V(allege)$sex <- atts[ ,4]

V(secrets)$family <- atts[ ,2]
V(secrets)$employed <- atts[ ,3]
V(secrets)$sex <- atts[ ,4]

Finally, save your work.

save(positive, file="data/positive.rda")
save(negative, file="data/negative.rda")
save(allege, file="data/allege.rda")
save(secrets, file="data/secrets.rda")
Node
    Categorical
        Color
        Shape
        Label size
        Layout

    Continuous
        Size
        Color
        Layout
        Label (Color or Size)

Tie
    Categorical
        Color
        Label
        Line type (solid, dotted, dashed, curved)

    Continuous
        Size
        Arrow size
        Label (value, size, color)

Emphasize and compare

Removing or hiding
    Removing certain nodes
    Minimize all but a few labels

Compare networks
    Save coordinates

Larger and denser networks will be more difficult to visualize

Saving High Quality Work.

If you go to the trouble to create good work. Then spend some effort at saving it in high quality format.

Screenshots are garbage and should be treated as such. Instead, save your work as a high quality jpeg, png, pdf, or similar. It is worth the extra effort.

jpeg("plot_name.jpg",      # Name the plot
     width=5, height=5,    # Set the size of the plot area
     units="in", res=400)  # Here, we are using inches. You can switch
                           # to "cm" if you prefer. "res" is resolution
                           # 400 is ok... prob higher res
                  

plot.igraph(network,           # add your network here
            vertex.label.cex=0.5)

dev.off()                  # Don't forget this part. 
                           # It ends the process and saves your work.

For example, see below:

jpeg("The Trial Plot.jpg",
     width=7, height=7,
     units="in", res=400)
par(mfrow=c(2,2), mar=c(0.5,0,2,0))
# make a two row, two column table of plots. margins (bottom, left, top, right) (default is 1 1 2 1)
plot(positive, main="Positive Sentiment", layout=coords)
plot(negative, main="Negative Sentiment", layout=coords)
plot(allege, main="Made Allegations About", layout=coords)
plot(secrets, main="Knows a Secret About", layout=coords)
par(mfrow=c(1,1))
#make the plot window plot just one graph at a time

dev.off()
## quartz_off_screen 
##                 2