+ - 0:00:00
Notes for current slide
Notes for next slide

R For SNA

Workshop 1

Eric Brewe
Associate Professor of Physics at Drexel University

4 August 2020, last update: 2020-08-11

1 / 50

Why shouldn't I just use Excel?

  • R is a programming language
    • Data live separate from analysis (this is good)
    • Data are imported, manipulated, represented, but not changed.
    • This means you can't screw up!
2 / 50

Why shouldn't I just use Excel?

  • R is a programming language
    • Data live separate from analysis (this is good)
    • Data are imported, manipulated, represented, but not changed.
    • This means you can't screw up!

Or at least it is hard to screw up

🤷

2 / 50

R is good for...

  • Data cleaning
  • Plotting
  • Summarizing
  • Manipulating data
  • Reproducibility
  • Sharing Code
3 / 50

Foundations of Network Analysis

4 / 50

What is a network?

  • Collection of object-like things that are connected.
    • Nodes/actors = Object-like things (Nouns)
      • Students in a class, Words in a novel, Banks...
      • Nodes can have attributes
        • Gender,
        • Word-type,
        • Market capitialization

NetworkImage

5 / 50

What is a network?

  • Collection of object-like things that are connected.
    • Ties/Links/Edges = Connections between nodes (Verbs)
      • Talked to each other, Are neigbors, Lend money, Sent text message...
        • Directional
        • Multiplex
        • Weighted

NetworkImage

6 / 50

Network Analysis is for the analysis of relational data

There are four basic assumptions:

  1. Nodes and interactions are interdependent*
  2. Edges allow flow between nodes
  3. Network models on indiviuals both constrain and provide opportunity for action
  4. Network models conceptualize structure as representation of lasting patterns of relations between actors

* Violates basic assumption of inferential statistics

Wasserman, S., Faust, K. (1994). Social network analysis: Methods and applications (Vol. 8). Cambridge university press.

7 / 50

What can we do with it?

Ego-Level Analyses

  • What can we know about the network of one person?
    • Ego density
    • Number of neighbors
    • Number of connected neighbors

8 / 50

What can we do with it?

Node-Level Analyses

  • What can we know about the position of people in a network?
    • Degree (In/Out/Total)
    • Geodesic Distance (Kevin Bacon)
    • PageRank
    • Target Entropy

9 / 50

What can we do with it?

Whole Network Analyses

  • What can we say about a whole network?
    • Density, Average path length, Giant component
    • Clustering
    • Homophily
    • Modeling
      • Block models
      • Small worldness

10 / 50

Historical Foundations

Joseph Moreno & Helen Hall Jennings (1932)

  • Established foundations of SNA

Quantitative Sociology/Anthropology

  • Davis Southern Women's Club (1941)
  • Small World Problem (1967)
  • Zachary's Karate Club (1977)

Seminal Articles

  • Milgram, Stanley "The small world problem" Psychology Today 2:1, (1967)
  • Grannovetter, Mark S. "The strength of weak ties" American Journal of Sociology (1973)

11 / 50

Modern Foundations of Network Analysis

Sociophysics (1990s)

  • Graph theory
  • Information theory
  • Computing
  • Used to study
    • Internet
    • Power grid
    • Transportation networks

Seminal Articles

  • Watts & Strogratz "Collective dynamics of small world networks" Nature (1998)
  • Page, Brin, Motwani, & Winograd "The PageRank citation ranking: Bringing order to the web" Stanford InfoLab (1999)
12 / 50

Important Takeaways from History

Two main camps

13 / 50

Important Takeaways from History

Two main camps

Statistical -> hypothesis testing

13 / 50

Important Takeaways from History

Two main camps

Statistical -> hypothesis testing

Graph theoretic -> network models and simulation

13 / 50

Important Takeaways from History

Two main camps

Statistical -> hypothesis testing

Graph theoretic -> network models and simulation

They often don't agree.

There is often distain.

They have different language, journals, conferences

13 / 50

Network Data in R

Sociomatrix/Adjacency Matrix

## 5 x 5 sparse Matrix of class "dgCMatrix"
##               
## [1,] . 1 1 1 1
## [2,] 1 . . . .
## [3,] 1 . . . .
## [4,] 1 . . . 1
## [5,] 1 . . 1 .
14 / 50

Network Data in R

Edgelist

## [[1]]
## + 4/5 edges from 184c722:
## [1] 1--2 1--3 1--4 1--5
## 
## [[2]]
## + 1/5 edge from 184c722:
## [1] 1--2
## 
## [[3]]
## + 1/5 edge from 184c722:
## [1] 1--3
## 
## [[4]]
## + 2/5 edges from 184c722:
## [1] 1--4 4--5
## 
## [[5]]
## + 2/5 edges from 184c722:
## [1] 1--5 4--5
15 / 50

What does this mean in terms of learning R?

  • We need to know something about the different types of data!

Data types (at least some of them)

  • Logical (T/F, 1/0)
  • Integers (whole numbers)
  • Numeric (numbers with decimal places)
  • Complex (I never use these)

Data storage (at least some of them)

  • Vectors = long columns of data (can be any type, but only one type of data)
  • Dataframes = like Excel pages (columns can hold different types of data)
  • Matrices = like dataframes, but they have named rows/columns
    • Adjacency Matrices are of type matrix.
  • Lists = the junkdrawer, can hold any type of data (including dataframes or matrices)
    • in igraph, networks are stored as lists.
16 / 50

What you need to know about R?

  • R is the programming language.
  • RStudio is the Integrated Development Environment (IDE)
  • Packages: groups of functions that are developed as open source
  • Base R
    • The group of packages preloaded into R
  • Tidyverse
    • Family of packages that are designed with the theory that programming should be readable by humans.
  • igraph
    • Package that is very useful for doing network analyses

Lets do this!

17 / 50

Using R and RStudio

  • Open RStudio (this will automatically open R)
  • Navigate to your folder titled "RForSNA"
  • In RStudio -> File -> New Project
    • Select "Existing Directory" (Unless you know Git)
  • In RStudio -> File -> New File -> RMarkdown
  • Save this file in the folder titled "RForSNA"

18 / 50

Lets Take a Tour of RStudio IDE

19 / 50

Let's Install Some Packages

You'll only need to do this once.

In Console

To install tidyverse package...

install.packages("tidyverse")

Repeat this with the following packages:

igraph tidygraph here ggraph

20 / 50

Let's Load Some Packages

You'll need to do this every time you restart R.

In Console

To load tidyverse package...

library(tidyverse) #tools for cleaning data 
library(igraph)  #package for doing network analysis
library(tidygraph) #tools for doing tidy networks
library(here) #tools for project-based workflow
library(ggraph) #plotting tools for networks
21 / 50

Let's Load Some Packages

You'll need to do this every time you restart R.

In Console

To load tidyverse package...

library(tidyverse) #tools for cleaning data 
library(igraph)  #package for doing network analysis
library(tidygraph) #tools for doing tidy networks
library(here) #tools for project-based workflow
library(ggraph) #plotting tools for networks

Once you have done this, you will want to put include a code chunk with all of your libraries into your markdown document so that you don't have to type this every time.

21 / 50

Let's get data into R.

I've sent you a csv file that includes the data for workshop 1, I hope you saved this in your folder titled "data".

If you have loaded the package "here" this should just work. If you have not loaded the "here" package you will need to set the working directory.

Again, you will want to include this as a code chunk in your RMD file. 

#This loads the csv and saves it as a dataframe titled WorkshopData
WorkshopData <- read_csv(here("data", "AnonSurveyData.csv"))
22 / 50

Let's have a look at the data

glimpse(WorkshopData)
## Rows: 34
## Columns: 18
## $ ID                      <dbl> 5106, 6633, 7599, 4425, 2495, 6355, 8810, 387…
## $ StartDate               <dttm> 2020-07-30 12:15:20, 2020-07-30 12:18:50, 20…
## $ EndDate                 <dttm> 2020-07-30 12:18:59, 2020-07-30 12:20:21, 20…
## $ Status                  <chr> "IP Address", "IP Address", "IP Address", "IP…
## $ Progress                <dbl> 100, 100, 100, 100, 100, 100, 100, 100, 100, …
## $ `Duration (in seconds)` <dbl> 219, 91, 137, 97, 127, 233, 97, 144, 103, 231…
## $ Finished                <lgl> TRUE, TRUE, TRUE, TRUE, TRUE, TRUE, TRUE, TRU…
## $ RecordedDate            <dttm> 2020-07-30 12:18:59, 2020-07-30 12:20:22, 20…
## $ SurveyID                <chr> "R_ssTGHZwy5EpQaNX", "R_2fv9VCk0tjdrDOr", "R_…
## $ ExternalReference       <lgl> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, N…
## $ DistributionChannel     <chr> "email", "email", "email", "email", "email", …
## $ UserLanguage            <chr> "EN", "EN", "EN", "EN", "EN", "EN", "EN", "EN…
## $ Q2                      <chr> "morning person.", "morning person.", "night …
## $ Q3                      <chr> "Brownies", "I don't care for dessert.", "Ice…
## $ Q4                      <chr> "Coffee,Water,The tears of my enemies", "Coff…
## $ Q5                      <dbl> 350, 12, 300, 0, 264, 4, 289, 550, 349, 300, …
## $ Q6                      <dbl> 350, 56, 5000, 50, 286, 250, 76, 185, 108, 22…
## $ `as.character(Q5)`      <dbl> 350, 12, 300, 0, 264, 4, 289, 550, 349, 300, …
23 / 50

Let's start cleaning up.

First, we don't need most of that data

There is a ton of data there that doesn't make sense for us to keep around.

We will use the '%>%' (pipe) operator and the verb select

WorkshopData %>%
  select(ID,Q2:Q6) -> WorkshopData
glimpse(WorkshopData)
## Rows: 34
## Columns: 6
## $ ID <dbl> 5106, 6633, 7599, 4425, 2495, 6355, 8810, 3877, 1554, 7743, 8353, …
## $ Q2 <chr> "morning person.", "morning person.", "night owl.", "morning perso…
## $ Q3 <chr> "Brownies", "I don't care for dessert.", "Ice Cream", "I don't car…
## $ Q4 <chr> "Coffee,Water,The tears of my enemies", "Coffee,Tea,Water,Milk", "…
## $ Q5 <dbl> 350, 12, 300, 0, 264, 4, 289, 550, 349, 300, 424, 426, 231, 290, 1…
## $ Q6 <dbl> 350, 56, 5000, 50, 286, 250, 76, 185, 108, 220, 500, 350, 412, 113…
24 / 50

Note the data are not numbers.

Let's start cleaning up.

Now we should actually take a look at the data

WorkshopData %>%
  head()
## # A tibble: 6 x 6
##      ID Q2            Q3                  Q4                            Q5    Q6
##   <dbl> <chr>         <chr>               <chr>                      <dbl> <dbl>
## 1  5106 morning pers… Brownies            Coffee,Water,The tears of…   350   350
## 2  6633 morning pers… I don't care for d… Coffee,Tea,Water,Milk         12    56
## 3  7599 night owl.    Ice Cream           Fruit Juice,Tea,Water,Fiz…   300  5000
## 4  4425 morning pers… I don't care for d… Fruit Juice,Coffee,Water       0    50
## 5  2495 morning pers… Brownies            Coffee,Water                 264   286
## 6  6355 night owl.    Ice Cream           Fruit Juice,Tea,Water          4   250
25 / 50

Let's check out the power of R

I am going to blast through these next slides, to show you some of the things that you might want to do with R

26 / 50

Let's start visualizing the data

For categorical data, you might want to get some counts.

Here is code to do this for the question about morning or night person.

WorkshopData %>%
  select(Q2) %>%
  group_by(Q2) %>%
  tally()
## # A tibble: 2 x 2
##   Q2                  n
##   <chr>           <int>
## 1 morning person.    19
## 2 night owl.         15
27 / 50

Let's start visualizing the data

For categorical data, you might want to get a histogram.

Here is code to do this for the favorite dessert type.

WorkshopData %>%
  select(Q3) %>%
  group_by(Q3) %>% 
  tally()
## # A tibble: 5 x 2
##   Q3                                      n
##   <chr>                               <int>
## 1 Brown Butter Chocolate Chip Cookies     2
## 2 Brownies                                9
## 3 Cheese                                  3
## 4 I don't care for dessert.               3
## 5 Ice Cream                              17
WorkshopData %>%
  select(Q3) %>%
  ggplot(aes(y = Q3)) + 
  geom_bar()

28 / 50

Let's explore beverage choices

WorkshopData %>%
  select(ID, Q4) %>%
  head()
## # A tibble: 6 x 2
##      ID Q4                                  
##   <dbl> <chr>                               
## 1  5106 Coffee,Water,The tears of my enemies
## 2  6633 Coffee,Tea,Water,Milk               
## 3  7599 Fruit Juice,Tea,Water,Fizzy Water   
## 4  4425 Fruit Juice,Coffee,Water            
## 5  2495 Coffee,Water                        
## 6  6355 Fruit Juice,Tea,Water

Notice, these are not tidy data, more than one variable per line

29 / 50

Let's dummy code the beverage data

For tidy data we want one value per row

WorkshopData %>%
  select(ID, Q4) %>%
  separate_rows(Q4, sep = ",") %>%
  head(10)
## # A tibble: 10 x 2
##       ID Q4                     
##    <dbl> <chr>                  
##  1  5106 Coffee                 
##  2  5106 Water                  
##  3  5106 The tears of my enemies
##  4  6633 Coffee                 
##  5  6633 Tea                    
##  6  6633 Water                  
##  7  6633 Milk                   
##  8  7599 Fruit Juice            
##  9  7599 Tea                    
## 10  7599 Water

We can dummy code these

WorkshopData %>%
  select(ID, Q4) %>%
  separate_rows(Q4, sep = ",") %>%
  mutate(Checked = 1) %>%
  pivot_wider(names_from = Q4,
              values_from = Checked,
              values_fill = 0)
## # A tibble: 34 x 11
##       ID Coffee Water `The tears of m…   Tea  Milk `Fruit Juice` `Fizzy Water`
##    <dbl>  <dbl> <dbl>            <dbl> <dbl> <dbl>         <dbl>         <dbl>
##  1  5106      1     1                1     0     0             0             0
##  2  6633      1     1                0     1     1             0             0
##  3  7599      0     1                0     1     0             1             1
##  4  4425      1     1                0     0     0             1             0
##  5  2495      1     1                0     0     0             0             0
##  6  6355      0     1                0     1     0             1             0
##  7  8810      1     0                1     0     0             0             1
##  8  3877      0     1                1     1     1             0             1
##  9  1554      0     1                1     0     0             1             0
## 10  7743      0     1                0     1     1             1             1
## # … with 24 more rows, and 3 more variables: `A delicious 12 year single malt
## #   scotch from the Scottish lowlands with notes of apple` <dbl>, `
## #   cinnamon` <dbl>, ` and dried fruit served with a single ice cube` <dbl>
30 / 50

Let's look at some quantitative data

First, let's summarize the reading data.

WorkshopData %>%
  select(Q5) %>%
  summarize(Ave = mean(Q5, na.rm = TRUE), 
            SD = sd(Q5, na.rm = TRUE))
## # A tibble: 1 x 2
##     Ave    SD
##   <dbl> <dbl>
## 1  327.  247.
31 / 50

Let's investigate groups

Are morning people or night owls reading longer books?

WorkshopData %>%
  select(Q2, Q5) %>%
  group_by(Q2) %>%
  summarize(Ave = mean(Q5),
            SD = sd(Q5))
## # A tibble: 2 x 3
##   Q2                Ave    SD
##   <chr>           <dbl> <dbl>
## 1 morning person.  321   215.
## 2 night owl.       335.  289.

We might want to use a boxplot to display these data

WorkshopData %>%
  select(Q2, Q5) %>%
  ggplot(., aes(x = Q2, y = Q5)) +
  geom_boxplot()

32 / 50

Let's look at readers vs. blueberries

Is there a relationship between length of book and estimates on number of blueberries?

Could do a scatter plot

WorkshopData %>%
  select(Q5:Q6) %>%
  mutate(Q5 = as.numeric(Q5), Q6 = as.numeric(Q6)) %>%
  ggplot(aes(x = Q5, y = Q6)) +
  geom_point()

33 / 50

Let's look at readers vs. blueberries

Is there a relationship between length of book and estimates on number of blueberries?

Or you could do a linear model

summary(lm(Q6 ~ Q5, data = WorkshopData))
## 
## Call:
## lm(formula = Q6 ~ Q5, data = WorkshopData)
## 
## Residuals:
##    Min     1Q Median     3Q    Max 
## -365.6 -313.8 -179.1  -98.8 4563.5 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)
## (Intercept) 382.9414   249.7136   1.534    0.135
## Q5            0.1785     0.6126   0.291    0.773
## 
## Residual standard error: 867.6 on 32 degrees of freedom
## Multiple R-squared:  0.002647,    Adjusted R-squared:  -0.02852 
## F-statistic: 0.08492 on 1 and 32 DF,  p-value: 0.7726
34 / 50

Let's prep our data for SNA

We will need to prep two separate files...

  1. An edgelist
  2. A file of attributes of the nodes.
35 / 50

Let's make an edgelist

There is an issue here.

I wanted to make these data anonymous (so we don't know who likes scotch)

But to do that I had to make the edgelist for you.

So I sent you an edgelist as a csv. Sorry.

Check out the edgelist

EL = read_csv(here("data", "AnonEL.csv"))
head(EL)
## # A tibble: 6 x 2
##      ID Connections
##   <dbl>       <dbl>
## 1  5106        5106
## 2  6633        6196
## 3  7599        5462
## 4  4425        7743
## 5  2495        3940
## 6  6355        6355
36 / 50

Let's assemble our node attributes

Before we can convert our Edgelist to a network, we should add in the attributes.

We have several candidate attributes:

  • Morning vs. Night
  • Dessert Type
  • Pages in book

We will develop a separate dataframe for the attributes. 

WorkshopData %>%
  select(ID, Q2, Q3, Q5) -> AttributeDf
37 / 50

Let's assemble our node attributes

Experience tells me that when you try to add attributes that you often make a mistake where the number attributes don't match up well to the number of nodes...but lets see.

gr <- graph_from_data_frame(EL, directed = TRUE)
plot(gr)

gr = as_tbl_graph(gr)
38 / 50

Let's add some attributes

So actually the easiest way to add the attributes is to add them while you make the graph.

But that isn't as easy as it seems

gr %>%
  activate(nodes) %>%
  mutate(AMPM = AttributeDf$Q2)

But that isn't as easy as it seems...

Let's sort out these attributes.

The warning was: "Input AMPM must be size 60 or 1, not 34."

What this means is we need to take our attributes dataframe and make sure all the nodes are listed.

To do this we need to:

  • Compile a dataframe of all nodes listed in the graph
  • Use a join to add attributes to this dataframe
39 / 50

Let's sort out these attributes (pt 2)

#This will get a vector of all nodes
gr %>%
  activate(nodes) %>%
  as_tibble() %>%
  transmute(ID = name) %>%
  mutate(ID = as.numeric(ID))-> GrNodes
#Now we pull in the attributes using a left_join
NodeAttributes = left_join(GrNodes, AttributeDf, by = "ID")

You should inspect Node Attributes

40 / 50

Let's sort out these attributes (pt 3)

Inspect the head

head(NodeAttributes)
## # A tibble: 6 x 4
##      ID Q2              Q3                           Q5
##   <dbl> <chr>           <chr>                     <dbl>
## 1  5106 morning person. Brownies                    350
## 2  6633 morning person. I don't care for dessert.    12
## 3  7599 night owl.      Ice Cream                   300
## 4  4425 morning person. I don't care for dessert.     0
## 5  2495 morning person. Brownies                    264
## 6  6355 night owl.      Ice Cream                     4
41 / 50

Let's sort out these attributes (pt 4)

Inspect the tail

tail(NodeAttributes)
## # A tibble: 6 x 4
##      ID Q2    Q3       Q5
##   <dbl> <chr> <chr> <dbl>
## 1  7128 <NA>  <NA>     NA
## 2  1050 <NA>  <NA>     NA
## 3  3799 <NA>  <NA>     NA
## 4  1651 <NA>  <NA>     NA
## 5  8984 <NA>  <NA>     NA
## 6  1958 <NA>  <NA>     NA
42 / 50

Let's finally assemble this graph.

gr %>%
  as_tbl_graph() %>%
  activate(nodes) %>%
  mutate(AMPM = NodeAttributes$Q2) %>%
  mutate(Dessert = NodeAttributes$Q3) %>%
  mutate(Pages = NodeAttributes$Q5) -> gr
summary(gr)
## IGRAPH 1723efe DN-- 60 101 -- 
## + attr: name (v/c), AMPM (v/c), Dessert (v/c), Pages (v/n)
43 / 50

Let's finally assemble this graph.

Take a look at your graph

## # A tbl_graph: 60 nodes and 101 edges
## #
## # A directed multigraph with 8 components
## #
## # Node Data: 60 x 4 (active)
##   name  AMPM            Dessert                   Pages
##   <chr> <chr>           <chr>                     <dbl>
## 1 5106  morning person. Brownies                    350
## 2 6633  morning person. I don't care for dessert.    12
## 3 7599  night owl.      Ice Cream                   300
## 4 4425  morning person. I don't care for dessert.     0
## 5 2495  morning person. Brownies                    264
## 6 6355  night owl.      Ice Cream                     4
## # … with 54 more rows
## #
## # Edge Data: 101 x 2
##    from    to
##   <int> <int>
## 1     1     1
## 2     2    35
## 3     3    14
## # … with 98 more rows
44 / 50

Let's make our plots look a bit better

We can add various elements to our plot

  • Color (good for grouping)
  • Shape (good for grouping)
  • Size (good for numeric)
  • Text (good for labels)
  • Layout

This is the most basic plot

```r
ggraph(gr) +
  geom_edge_link() +
  geom_node_point()
```

Not super pretty

45 / 50

Let's make our plots look a bit better

We can add various elements to our plot

  • Layout, there are lots of options -Try circle?
```r
ggraph(gr, layout = 'circle') +
  geom_edge_link() +
  geom_node_point()
```

46 / 50

Let's make our plots look a bit better

We can add various elements to our plot

  • Shape, we can make night owls one shape and morning people a differnt shape...
```r
ggraph(gr, layout = 'circle') +
  geom_edge_link() +
  geom_node_point(aes(shape = AMPM))
```

Not great

47 / 50

Let's make our plots look a bit better

We can add various elements to our plot

  • Color, lets use the dessert type to define a color
```r
ggraph(gr, layout = 'circle') +
  geom_edge_link() +
  geom_node_point(aes(color = Dessert))
```

48 / 50

Let's make our plots look a bit better

We can add various elements to our plot

  • Size, lets make the nodes different sizes based on the number of pages in the last book read.
```r
ggraph(gr, layout = 'circle') +
  geom_edge_link() +
  geom_node_point(aes(size = Pages))
```

49 / 50

Let's make our plots look a bit better

We can add various elements to our plot

  • Let's jam them all together!
```r
ggraph(gr, layout = 'circle') +
  geom_edge_link() +
  geom_node_point(aes(shape = AMPM, 
                      color = Dessert, 
                      size = Pages))
```

50 / 50

Why shouldn't I just use Excel?

  • R is a programming language
    • Data live separate from analysis (this is good)
    • Data are imported, manipulated, represented, but not changed.
    • This means you can't screw up!
2 / 50
Paused

Help

Keyboard shortcuts

↑, ←, Pg Up, k Go to previous slide
↓, →, Pg Dn, Space, j Go to next slide
Home Go to first slide
End Go to last slide
Number + Return Go to specific slide
b / m / f Toggle blackout / mirrored / fullscreen mode
c Clone slideshow
p Toggle presenter mode
t Restart the presentation timer
?, h Toggle this help
Esc Back to slideshow