Stat 437 Lecture Notes 2d

Xiongzhi Chen

Washington State University

Other visualization techniques

Overview

We will cover

Visualizing data with geographical information
Visualizing network data
Visualizing tree-like structures
Dynamic visualization

The contents are gathered from online resources.

R packages Needed

The R packages we will discuss:

ggmap: provide maps for visualization
igraph: produce graphs
ggdendro: produce dendrograms
Plotly: create interactive web-based plots
gganimate: create dynamic plots

Visualization via `ggmap`

Overview

Some data (e.g., car accidents, insurance claims, transmission of infectious disease) often contain geographical information, and visualizing such data over a map can be quite informative.

The R package ggmap allows an overlay of geometric objects on maps, where locations are identified by their latitudes and longitudes. Details on how to use ggmap can be found at https://github.com/dkahle/ggmap
ggmap uses Google maps. However, Google has recently changed its API requirements, and ggmap users are now required to register with Google (and possibly pay a fee for using Google’s maps beyond a quota)

Illustration: background

This example is borrowed from https://blog.dominodatalab.com/geographic-visualization-with-rs-ggmaps/.

The dataset can be downloaded from https://app.dominodatalab.com/u/seanlorenz/ggmap-demo/view/data/vehicle-accidents.csv. It contains accidents caused by motor vehicles MV.Number and by non-motor vehicles NM.Number for 50 US states.

We will visualize MV.Number for 48 states (excluding Alaska and Hawaii) via ggmap.

Data set at a glance

Read data from local drive; make sure variables have correct object types
Remove observations for Alaska and Hawaii (due to US map properties)

> mydata = read.csv("vehicle-accidents.csv")
> mydata$State <- as.character(mydata$State)
> mydata$MV.Number = as.numeric(mydata$MV.Number)
> library(dplyr)
> mydata = mydata %>% 
+   filter(State !="Alaska" & State !="Hawaii")
> head(mydata %>% select(State,MV.Number))
        State MV.Number
1     Alabama        78
2     Arizona       145
3    Arkansas        46
4  California       722
5    Colorado        77
6 Connecticut        40

Obtain latitudes and longitudes

Load library ggmap
Obtain latitudes and longitudes for each state in data set (registration with Google is needed for this purpose)
Create a data frame that has the latitudes and longitudes

> library(ggmap)
> for (i in 1:nrow(mydata)) {
+   latlon = geocode(mydata[i,1])
+   mydata$lon[i] = as.numeric(latlon[1])
+   mydata$lat[i] = as.numeric(latlon[2])
+ }
> mv_num_collisions = data.frame(mydata$MV.Number,
+               mydata$lon,mydata$lat,mydata$State)
> colnames(mv_num_collisions) = c('MV.Number',
+               'lon','lat','State')

View latitudes and longitudes

Longitudes (lon) and latitudes (lat) for some states:

  MV.Number        lon      lat       State
1        78  -86.75026 32.75041     Alabama
2       145 -111.50098 34.50030     Arizona
3        46  -91.55780 33.98174    Arkansas
4       722 -118.02135 35.12562  California
5        77 -105.50083 39.00027    Colorado
6        40  -72.66648 41.66704 Connecticut

Obtain US map and plot data

Obtain US map by specifying map range and map type
circle_scale_amt sets scaling parameter for the circles to be overlayed on map
scale_size_continuous(range=NULL) specifies the minimum and maximum sizes of plotting symbol (“circle” in this example) after transformations
labs(x = "",y=""): no labels for axes

> us <- c(left = -125, bottom = 25.75, right = -67, top = 49)
> map <- get_stamenmap(us, zoom = 5, maptype = "toner-lite")
> circle_scale_amt <- 0.05
> ggmap(map)+geom_point(aes(x=lon,y=lat),
+  data=mv_num_collisions,col="orange", alpha=0.4,
+  size=mv_num_collisions$MV.Number*circle_scale_amt)+
+  scale_size_continuous(range=
+                   range(mv_num_collisions$MV.Number))+
+   labs(x = "",y="")

Visualization with overlay

The larger the circle, the higher the accident rate/number
East coast states more accidents than northwest states

Visualizing with `igraph`

Overview

Binary relationships, such as interactions between two individuals, a disease transmitting between two subjects, and reaction between two chemicals, among subjects or objects can be represented as networks (and equivalently as graphs).

Each “subject” in a network is represented by a “node” or “vertex” in a graph
Each “interaction” or “relationship” between two subjects or objects in a network is represented by an “edge” between their associated nodes in a graph

Overview

For networks or graphs, we often are interested in

if a network or graph can be split into smaller sub-networks or sub-graphs such that on average there are more connections between the nodes in each sub-network than there are connections between sub-networks
if a node in a graph has many more connections with other nodes than does any other node.

The R package igraph creates graphs as representations of networks.

Illustration: background

Materials to be presented are adopted from http://kateto.net/network-visualization.

The data set contains relationships among several medias. Among the medias, if one is referenced by another, either by mention or hyperlink, then the corresponding two medias are “connected” and there is an edge between them
Relationships among the medias will be represented as a graph, where a media is a “node”, and a relationship (i.e., reference by mention or hyperlink) is an “edge”

A look at the dataset

Load “nodes” and “links” (as edges):

> nodes <- read.csv("Dataset1-Media-Example-NODES.csv", 
+                   header=T, as.is=T)
> links <- read.csv("Dataset1-Media-Example-EDGES.csv", 
+                   header=T, as.is=T)

> head(nodes)
   id               media media.type type.label audience.size
1 s01            NY Times          1  Newspaper            20
2 s02     Washington Post          1  Newspaper            25
3 s03 Wall Street Journal          1  Newspaper            30
4 s04           USA Today          1  Newspaper            32
5 s05            LA Times          1  Newspaper            20
6 s06       New York Post          1  Newspaper            50
> head(links)
  from  to      type weight
1  s01 s02 hyperlink     22
2  s01 s03 hyperlink     22
3  s01 s04 hyperlink     21
4  s01 s15   mention     20
5  s02 s01 hyperlink     23
6  s02 s03 hyperlink     21

Create an `igraph` object

directed=T creates a directed graph (to emphasize “media A” is referenced by “media B”)
Note d=links,vertices=nodes and “media” in net

> library(igraph)
> net=graph_from_data_frame(d=links,vertices=nodes,directed=T)
> net
IGRAPH d4cf23c DNW- 17 49 -- 
+ attr: name (v/c), media (v/c), media.type (v/n),
| type.label (v/c), audience.size (v/n), type (e/c),
| weight (e/n)
+ edges from d4cf23c (vertex names):
 [1] s01->s02 s01->s03 s01->s04 s01->s15 s02->s01 s02->s03
 [7] s02->s09 s02->s10 s03->s01 s03->s04 s03->s05 s03->s08
[13] s03->s10 s03->s11 s03->s12 s04->s03 s04->s06 s04->s11
[19] s04->s12 s04->s17 s05->s01 s05->s02 s05->s09 s05->s15
[25] s06->s06 s06->s16 s06->s17 s07->s03 s07->s08 s07->s10
[31] s07->s14 s08->s03 s08->s07 s08->s09 s09->s10 s10->s03
+ ... omitted several edges

Create a graph

Two sub-networks; Wall Street Journal (has the most edges)

> # remove multiple edges and loops
> net1=simplify(net,remove.multiple=T,remove.loops=T)
> # add "vertex.label" and specify label color
> plot(net1, edge.arrow.size=.4, edge.color="blue",
+     vertex.label=V(net)$media, vertex.label.color="black")

Visualization via `ggdendro`

Overview

[From wiki] A dendrogram is a diagram representing a tree.

In hierarchical clustering, it illustrates the arrangement of the clusters produced by some analyses
In decision making, it illustrates the decisions to be made under different situations
In phylogenetics, it displays the evolutionary relationships among various biological taxa. In this case, the dendrogram is also called a “phylogenetic tree”

The R package ggdendro creates dendrograms.

Illustration: background

The USArrests data set (included in the library ‘ggdendro’) contains arrests per 100,000 residents in each of the 50 US states in 1973:

Arrests for 3 types of crimes: Murder, Assault, Rape
UrbanPop (percent of population living in urban areas)

> library(ggplot2); library(ggdendro); head(USArrests)
           Murder Assault UrbanPop Rape
Alabama      13.2     236       58 21.2
Alaska       10.0     263       48 44.5
Arizona       8.1     294       80 31.0
Arkansas      8.8     190       50 19.5
California    9.0     276       91 40.6
Colorado      7.9     204       78 38.7

We will classify (or cluster) states based on the 4 features (and the Euclidean distance).

Create a dendrogram

Dendrogram suggests 2 large clusters or 4 smaller clusters

> hc <- hclust(dist(USArrests), "ave")  # clustering
> ggdendrogram(hc, rotate = TRUE) # rotate by 90 degrees

License and session Information

License

> sessionInfo()
R version 3.5.0 (2018-04-23)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 10 x64 (build 18362)

Matrix products: default

locale:
[1] LC_COLLATE=English_United States.1252 
[2] LC_CTYPE=English_United States.1252   
[3] LC_MONETARY=English_United States.1252
[4] LC_NUMERIC=C                          
[5] LC_TIME=English_United States.1252    

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods  
[7] base     

other attached packages:
[1] knitr_1.29

loaded via a namespace (and not attached):
 [1] compiler_3.5.0  magrittr_1.5    tools_3.5.0    
 [4] htmltools_0.5.0 revealjs_0.9    yaml_2.2.1     
 [7] stringi_1.4.6   rmarkdown_1.11  highr_0.8      
[10] stringr_1.4.0   xfun_0.15       digest_0.6.25  
[13] rlang_0.4.6     evaluate_0.14

Stat 437 Lecture Notes 2d

Xiongzhi Chen

Washington State University

Other visualization techniques

Overview

R packages Needed

Visualization via ggmap

Overview

Illustration: background

Data set at a glance

Obtain latitudes and longitudes

View latitudes and longitudes

Obtain US map and plot data

Visualization with overlay

Visualizing with igraph

Overview

Overview

Illustration: background

A look at the dataset

Create an igraph object

Create a graph

Visualization via ggdendro

Overview

Illustration: background

Create a dendrogram

License and session Information

Visualization via `ggmap`

Visualizing with `igraph`

Create an `igraph` object

Visualization via `ggdendro`