Social Network Analysis in R

Laboratory One
Benjamin Lind

Social Network Analysis: Internet Research
St. Petersburg
18 August, 2013 (11:45-13:15)

About R

Implements the S programming language
Open source
Object oriented programming language
Base & user-contributed packages
Meets practically all statistical purposes
Availability: r-project.org


Start R!

Type "R" into your terminal

Basics

  • Command line-only interface
    • Reliable, Replicable, Records
  • GUI options do exist, but...
    • outside of scope for our purposes
  • Assignment. Functionally equivalent methods
  • x <- 1 #Preference: primary method
  • x = 1 #Preference: within functions
  • 1 -> x #Preference: never 
  • Once assigned, x becomes an "object"
  • ls() 

Case Sensitivity

Try

Cheburashka <- "a cartoon character"Cheburashka
cheburashka

What does the error mean?

Hard Drive Navigation

  • R runs within a folder on your hard drive
    • Location where R reads & writes
  • Which folder are you in now?

orig.folder<-getwd()orig.folder
dir.create("new.folder")
list.files()
setwd("new.folder"); list.files()
setwd(orig.folder)
file.remove("new.folder"); list.files()

Packages

  • Preloaded (e.g., basestats)
  • Dependencies
  • User-contributed
    • Online research
      • RCurl
      • rjson
      • googleVis
      • twitteR
      • ...
    • Social network analysis
      • statnet
      • igraph
      • tnet
      • ...

Packages

To install a package
 install.packages("igraph")

To load a package
 library(igraph)

To unload a package
 detach("package:igraph", unload=TRUE)

Missing Data

Typically coded as "NA".  For example
x <- NA
Missing codes
NA #Non-applicable, missing
NULL #Undefined or empty value
NaN #Not a Number
Inf #Infinite 
Indicated by functions
is.na(x)
is.null(x)
is.nan(x)
is.infinite(x)

Data Types

Storing and accessing information within R

Picture by KENPEI

Data Types

  • Elements
    • Numeric
    • Character
      • Use double quotes
    • Logical
      • TRUEFALSETF10
  • Objects
    • Vectors
    • Matrices
    • Arrays
    • Data frames
    • Lists
    • Closures

Elements

Identification
is.numeric(42)
is.numeric("42")
is.numeric(3.12)
is.logical(T)
is.logical(FALSE)
is.character(FALSE)
is.character("42")
is.character("HSE Network Workshop") 
Conversion
as.numeric("42") + 1
as.character(as.numeric("0") + 1)
as.character(!as.logical("1"))
as.character(as.logical(as.numeric("0") + 1))
as.character(is.logical(as.logical(as.numeric("0")) + 1))
as.character(!as.logical(as.numeric("1")))  

Object Types: Vectors

Vectors collect elements in one dimension
(Photo by Frank C. Müller)

Object Types: Vectors

Vectors collect elements in one dimension
RU.cit<-c("Mos", "StP", "Nov", "Yek") #Create vector
RU.cit<-c(RU.cit, "Niz") #Add element
RU.cit[3] #Retrieve element 3
RU.pops<-c(10.4, 4.6, 1.4, 1.3) #Create vector
names(RU.pops)<-RU.cit[c(1,2,3,4)] #Name vector
RU.pops[c(4,2,3,1)] #Permute the order and retrieve
length(RU.pops) #Retrieve the length
RU.vis<-c(TRUE, TRUE, FALSE, FALSE) #Create vector
RU.vis[-c(4,2)] #Remove elements 4 and 2
Remaining object types build upon this logic

Object Types: Vectors


Sequences
all(1:5 == c(1, 2, 3, 4, 5))
all(1:5 == seq(from=1, to=5, by=1)) 

Repetitions
print(rep(1, times=5))
print(rep(NA, times=4))
print(rep("Beetlejuice", times=3)) 

Object Types: Matrices and Arrays














(Picture by David.Asch)

Object Types: Matrices & Arrays

Matrices
  • Two dimensions:
    • nrow(); ncol() 
  • Assignment:
    • x[1,5]<-1; x[2,]<-x[1,5]>0 
  • More columns, rows:
    • cbind(); rbind()
  • Names: rownames(); colnames()
Arrays
  • Three dimensions:
    • dim(); length(x[1,1,])  
  • Assignment:
    • x[3,2,6]<-1; x[4,,2]<-NA  
  • Names: dimnames()

Object Types: Matrices

Turning vectors into matrices
RU.mat<-cbind(pop=RU.pops, vis=RU.vis)
rownames(RU.mat)<-RU.cit[c(1:4)]
colnames(RU.mat) 
Alternatively
RU.mat<-matrix(,nrow=length(RU.pops), ncol=2)
colnames(RU.mat) <- c("pop", "vis")
rownames(RU.mat) <- names(RU.pops)
RU.mat[,"pop"] <- RU.pops
RU.mat[,"vis"] <- RU.vis 
Try
print(RU.mat[1,])print(RU.mat[,"pop"])
print(RU.mat[1,2])print(RU.mat[1,,]) #What does the error mean?

Object Types: Data Frames

Data frames are like matrices, but allow multiple element types. Compare:
print(RU.mat[,"vis"])
print(RU.vis) 
To retain initial element type
RU.df<-as.data.frame(RU.mat)
RU.df$vis<-as.logical(RU.df$vis) 
These commands are equivalent statements
print(RU.df$vis[2])
print(RU.df[2,"vis"])
print(RU.df[["vis"]][2])
print(RU.df[[2]][2]) 

Object Types: Lists

Lists collect objects of (potentially) varying types & dimensions
RU.list<-list(RU.mat, RU.vis, RU.df)
names(RU.list)<-c("Mat","Vec","DF") 

These commands are equivalent statements
print(RU.list$DF$vis[2])
print(RU.list[["DF"]]$vis[2])
print(RU.list[[3]]$vis[2])
print(RU.list[["DF"]][["vis"]][2])
print(RU.list[["DF"]][2,"vis"])
print(RU.list[["DF"]][2,2]) 

Object Types: Closures

Closures are functions
Functions accept input (optional) & return output

(Photo by erix!)

Object Types: Closures

Closures are functions
Functions accept input (optional) & return output
xTimesxMinus1<-function(x)
  return(x*(x-1))xTimesxMinus1(42)

InvHypSineTrans<-function(x){
  x.ret<-log(x+sqrt(x^2 + 1))   return(x.ret)}

plot(x=1:1000, InvHypSineTrans(1:1000), main="Hyperbolic Sine Transformation", xlab="x", ylab="y")

Compiled functions
library(compiler)InvHypSineTrans.cmp<-cmpfun(InvHypSineTrans)system.time(InvHypSineTrans(1:10000000))system.time(InvHypSineTrans.cmp(1:10000000))

Help

How to overcome the inevitable hurdles

Picture from Richard Milnes

Help

If you're looking for a function, but don't know the name:
??"Chi Squared"
help.search("Chi Squared") 
Once you know the name of the function
?chisq.test 
Want to know how the function is written?
print(chisq.test) 
Still stumped?

Help

Exercise
  1. Find the randomized normal distribution function
  2. Use it to create two vectors of length 100
    1. One with mean of zero & standard deviation of one
      1. Call it x
    2. A second with mean of 50 & standard deviation of 25
      1. Call it y
  3. Find the linear regression model command
  4. Regress y on x

Apply Family


Apply Family

Apply administers a function across elements within an object
Better than for() loops in R

lab.seq<-seq(from=-100,to=100,by=20)
?sapply
lab.mat<-sapply(lab.seq, rnorm, n=100)
fix(lab.mat)
?apply
lab.means<-apply(lab.mat, 2, mean)
plot(x=lab.seq, y=lab.means) 

Apply Family: Exercise

Recreate the last exercise, but with the following modifications

  1. Use a sequence from 1 to 100 by 5
  2. Run the sapply() line, but...
  • Set the mean to zero
  • Apply the sequence to rnorm()'s standard deviation
  • Run the apply() line, but...
    • Retrieve the standard deviation instead of the mean

    Plot your results and show them to me

    Loading and Saving

    Reusing material between R sessions

    (Photo by mRio)

    Loading & Saving: Source

    Text files for executable R code

    Like 'do files' for Stata
    Write them in text editors, following R syntax
    Exercise
    In text editor, type
    test.val<-"It works!"
    test.val2<-paste(test.val, "Really")
    Save as "test.val.source.txt" in working directory
    In R, type
    source("test.val.source.txt")
    ls()
    print(test.val); print(test.val2) 

    Loading & Saving

    • History
      • RHistory files retain a log of your session
      • May be opened in a text editor
      • Should end in ".RHistory" extention
      • Usage
        savehistory("Filename.RHistory")
        loadhistory("Filename.RHistory") 
    • Images
      • Saves every object in the environment (i.e., ls())
      • Usage
        save.image("Filename.RData")
        load("Filename.RData") 

    SNA 2013-R Lab 1

    By Benjamin Lind

    SNA 2013-R Lab 1

    • 4,114
    Loading comments...

    More from Benjamin Lind