# Introduction to network data in R
# Author: Jake Fisher
# This first lab will focus on how to manage data. We will cover:
# - loading data & objects
# - common data mangement tasks using dplyr & tidyr
# - constructing network objects with statnet
# - save and recall network (incl. vertex and edge) attributes
# Time permitting, we will also discuss:
# - managing lists
# - functions
##### Loading data & objects #####
# R is an object-oriented language, which puts it somewhere between statistical
# computing languages, like SAS, Stata, or SPSS, and traditional programming
# languages like Python, Java, or C++. we will focus mainly on using R for
# statistical computing.
# Everything in R is saved as an object in your R workspace. For example,
# let's save the number 10 to an object called "a":
a <- 10
# Object names can contain letters, periods, underscores, and numbers. Objects
# can contain almost anything. Let's save a new object:
R.is.free <- "...but you get what you pay for"
# Now we've saved a character string to the object R.is.free . We can print
# the value of that object to the screen by typing the name of the object:
R.is.free
# And we can list the objects present in the workspace with ls()
ls()
# We can remove (delete) objects from the workspace with rm()
rm(a)
ls()
# It's usually wise to start your scripts off by clearing out the workspace.
# You do that with the following command:
rm(list = ls()) # Note that we can nest two commands
# For a typical data analysis exercise, we usually use a "data.frame". This
# works like a rectangular dataset that you might see in any other program.
df <- data.frame(x = 1:10, y = 11:20) # 1:10 creates a vector: 1, 2, ..., 10
df
# R holds everything in the workspace in memory (RAM). This is similar to
# Stata, and different from SAS, which reads in rows one at a time. Unlike
# Stata, R can hold multiple datasets in memory at once.
(df2 <- data.frame(a = letters[1:10], b = LETTERS[1:10]))
# (Aside: letters is a vector of "a", "b", ..., "z", and LETTERS is a vector of
# "A", "B", ..., "Z". We took the first 10 elements of each vector for this
# data.frame. Wrapping an assignment statement in parentheses will print the
# assigned object to the screen. E.g., (a <- "something") will print whatever
# is stored in a to the screen.)
# Both datasets are held in memory, which will make merges, etc., easy. It will
# also mean that we need to refer to a particular dataset when we try to work
# with the variables in that dataset.
ls()
# Most of the time, we will not enter data directly in our R script. Instead,
# we will read it in from an external file.
# It is often helpful to set the working directory, to tell R where to look for
# files.
setwd("C:/Users/fishe/Box Sync/Home Folder jcf26/SNH2017/Instructor dropbox") # Note: forward slashes
# If you want to check your working directory, you can use:
getwd()
# All other filenames will be relative to that location.
# R has functions to read in CSV files
dat <- read.csv("ahs_wpvar.csv", stringsAsFactors = FALSE)
# It can also read in Stata .dta files, and SAS .sas7bdat files using
# contributed packages (foreign, and sas7bdat, respectively).
# If you like to "look" at your data, Rstudio can open a spreadsheet viewer,
# but really it's easier to look in the console
head(dat) # only prints the first 6 rows
# We can also print just the column names using names(), the dimensions using
# dim(), or the number of rows or columns using nrow() or ncol()
names(dat)
dim(dat)
nrow(dat)
ncol(dat)
##### Data management with dplyr & tidyr #####
# Data management in base R is, frankly, a big headache. We will use a new,
# contributed package called dplyr to manage the data.
# If you haven't already, install dplyr:
# install.packages("dplyr")
# This only needs to be done once per computer you use R on.
# To use a contributed package, you have to load it using the library command.
# This must be done in every R script you run that uses the package.
library(dplyr)
# dplyr breaks data management into (about) 5 simple tasks: subsetting by rows,
# subsetting by columns, sorting, and creating new variables
# To subset by rows, we use the filter command. This creates a dataset with
# only smokers:
smokers <- filter(dat, PSMOKES == 1)
head(smokers) # note that PSMOKES is only 1's
# To subset by columns, we use the select command. This creates a dataset with
# only the columns for grade and smoker
grade.smoker <- select(dat, grade, PSMOKES)
head(grade.smoker) # Notice there are only 2 columns now
# To sort, we use arrange
sorted.grade <- arrange(grade.smoker, grade)
head(sorted.grade)
# To create new variables, we use mutate
hs.smokers <- mutate(grade.smoker, hs = grade > 8)
head(hs.smokers)
# dplyr also provides the %>% (read: "pipe") operator, which allows you to
# string these simple commands together. The pipe passes the output of the
# previous command to the first argument of the next command, or to a position
# that you specify with a period. E.g., if you're sorting the data, these are
# equivalent:
head(arrange(hs.smokers, desc(grade)))
hs.smokers %>% arrange(., desc(grade)) %>% head(.)
hs.smokers %>% arrange(desc(grade)) %>% head
# I will use the latter construction from now on. Let's recreate hs.smokers
# using chained commands:
hs.smokers <- dat %>%
select(grade, PSMOKES) %>%
mutate(hs = grade > 8)
# Recoding variables with mutate deserves extra attention, because it's really
# important. The structure of a mutate command is
# mutate(new_var1 = (vector1), new_var2 = (vector2), ...)
# So, to create a new variable, we have to create a vector with the new values
# to assign to it. In the case above, grade > 8 created a logical (i.e.,
# TRUE/FALSE) vector indicating whether each element of the column grade was
# greater than 8.
# The most common uses for this are:
# 1. calculating a new number from an existing number (or numbers),
# 2. creating categorical variables from numeric ones
# As an example of 1, let's calculate 3 variables: a logged variable, a squared
# variable, and an interaction of two numeric variables
# To modify a dataset "in place" -- meaning that you do not make a new object
# from the dataset, you can use the %<>% operator from the magrittr package.
# Install magrittr if you haven't already:
# install.packages("magrittr")
# Load magrittr
library(magrittr)
dat %<>%
mutate(idg.log = log(IDG_SS), idg.sq = IDG_SS ^ 2,
noms.times.grade = totnoms * grade)
head(dat)
# As an example of 2, let's split IDG_SS into 3 categories. To do that, we will
# split nominations into categories.
# First, an aside -- let's see what the distribution of IDG_SS looks like.
# To look at the distribution, we use ggplot2, a popular visualization package
# install.packages("ggplot2")
library(ggplot2)
qplot(x = IDG_SS, data = dat, geom = "histogram")
# Let's do 3 categories: 0 (isolate), 1 - 10 (average), 11 - max (popular).
# We'll use derivedFactor from the mosaic package for this.
# install.packages("mosaic")
library(mosaic)
dat %<>%
mutate(popularity = derivedFactor(
isolate = IDG_SS == 0,
average = IDG_SS < 11,
.default = "high",
.method = "first"))
# Note that derivedFactor is quite slow on large datasets. Other approaches to
# this include recode from the car package, or cut from base R.
# Let's check and make sure we did that right:
with(dat, table(popularity, IDG_SS))
# What does the distribution of this look like?
qplot(x = popularity, data = dat, geom = "bar")
# Fun with ggplot...
p <- qplot(x = popularity, y = IDG_SS, data = dat, geom = "boxplot") +
coord_flip()
p
p + scale_y_sqrt() # sqrt instead of log because of 0 values
p + theme_bw()
# dplyr also contains a number of other useful functions -- like the left_join,
# right_join, inner_join, etc., functions for merges -- that I will not cover
# here. Read the vignette for dplyr to get a better idea of how to use these
# functions.
vignette("introduction", package = "dplyr")
vignette(package = "dplyr") # view a list of all the vignettes for dplyr
# (Most packages have a vignette to explain how to use them. It's often worth
# reading.)
##### Creating a network object #####
# Following best practices, we will begin by setting up the workspace:
rm(list = ls())
setwd("C:/Users/fishe/Box Sync/Home Folder jcf26/SNH2017/Instructor dropbox") # Note: forward slashes
library(statnet) # lots of messages...
# Statnet can create a network object from any of the common network data
# structures: edgelists, adjacency lists, and adjacency matrices.
# Let's start with an example of an adjacency matrix. Here I'll define a
# network where a vertex is a member of my immediate family, and an edge
# indicates "parent":
people <- c("mom", "dad", "sister", "me")
(adj.mat <- matrix(0, nrow = 4, ncol = 4)) # create empty matrix
rownames(adj.mat) <- people
colnames(adj.mat) <- people
adj.mat["mom", "sister"] <- 1
adj.mat["mom", "me"] <- 1
adj.mat["dad", "sister"] <- 1
adj.mat["dad", "me"] <- 1
adj.mat # completed adjacency matrix
# We can create a network object using the network() function
(family <- network(adj.mat, matrix.type = "adjacency"))
# Network objects have a number of special functions for them. We will cover
# these in the next section. For now, let's plot the network
plot(family) # not terribly informative
# Note that we could have expressed the information above more compactly if we
# just listed the cell values that should be 1 (i.e., mom -> sister, mom -> me,
# etc.) and assumed everything else should be 0. That's called an edgelist.
# We can generate a network from an edgelist, too:
(edges <- data.frame(from = c(rep("mom", 2), rep("dad", 2)),
to = rep(c("sister", "me"), 2)))
family.2 <- network(edges, matrix.type = "edgelist")
# We can show the adjacency matrix from the network object using as.matrix
as.matrix(family.2) # order of the rows is different, but network is the same
# And finally, we could compress this even more by holding one the sender in
# one column, and the receiver in subsequent columns. This is called an
# adjacency list.
(adj.list <- data.frame(parent = c("mom", "dad"), child1 = rep("sister", 2),
child2 = rep("me", 2)))
# Statnet doesn't handle adjacency lists well, but they are the most common type
# of raw survey network data. Fortunately, we can convert from an adjacency
# list to an edgelist easily, with the tidyr package
library(tidyr)
library(dplyr) # for the pipe operator
library(magrittr) # for the %<>% operator
adj.list %>%
gather(key = "nomination.number", value = "alter", -parent)
# Gather strings your data out into a long column. We can select which
# variables to string out -- in this case, we wanted to keep the column for
# parent separate, and to repeat the values for each child.
# If you drop the middle column, you have an edgelist...
adj.list %>%
gather(key = "nomination.number", value = "alter", -parent) %>%
select(-nomination.number)
# ...which you can pass to network!
adj.list %>% # note that we didn't save the output
gather(key = "nomination.number", value = "alter", -parent) %>%
select(-nomination.number) %>%
network(., matrix.type = "edgelist")
# That's great, but how do we do it with a real dataset? Let's try with the
# faux Add Health data.
dat <- read.csv("ahs_wpvar.csv", stringsAsFactors = F)
# So what's in here, anyway?
names(dat)
# First of all, there are several communities in these data. Let's pick the
# first one:
cmty1 <- dat %>% filter(commcnt == 1)
# Notice that the "friend" variables contain "fnid", and no other variables do.
# We can use that to create an edgelist:
el <- cmty1 %>%
select(ego_nid, contains("fnid")) %>%
gather(key = "nomination", value = "alter", -ego_nid)
head(el) # hmmmm.... contains 99999 and NA values. Let's standardize those
el %>% arrange(ego_nid) %>% head
el %<>% mutate(alter = ifelse(alter == 99999, NA, alter))
# Now you would think this would work as a network...
el %>%
select(-nomination) %>%
network(., matrix.type = "edgelist")
# but statnet also can't really deal with missing values in the edgelist. So
# the best way to do this is to construct an empty network first, and then
# add the edges from the edgelist. This has the advantage of including isolates
# too.
# To do that, first we get a list of the people who are involved in the network:
# Get unique list of people in the network
(ah.people <- with(el, unique(c(ego_nid, alter)))) # this would be all the people, including those out of school
# ah.people <- unique(dat$ego_nid) # this is only people who took the survey
el %<>% filter(!is.na(alter)) # drop missing cases
# We're not going to do that, because it includes a lot of people who didn't
# take the survey.
# Now create an empty network
add.health <- network.initialize(n = length(ah.people))
# statnet adds edges by index values, not by names of variables. So, to make
# this work, we have to create an index value of 1, 2, ..., N, where N is the
# number of people in the network
el %<>%
mutate(ego.idx = match(ego_nid, ah.people),
alter.idx = match(alter, ah.people))
# Finally, add the edges from the edgelist
add.edges(add.health, el$ego.idx, el$alter.idx)
# Now, as a quick gut check, let's plot it:
set.seed(919)
plot(add.health)
# (As an aside, note that you could create a list of all the networks using
# lapply...)
##### Save and recall network attributes #####
# One of the benefits of statnet is that it's possible to save attributes for
# different pieces of the network. We can save vertex attributes, like whether
# the person is a smoker, edge attributes, like whether the ties is homophilous,
# or network attributes, like the name of the school, or the survey wave. The
# procedure for this is simple; let's illustrate a few.
# First, network attributes are characteristics of the network as whole. Set
# them with %n%
add.health %n% "cmty" <- 1
add.health # note "cmty" in there
# Second, vertex attributes are characteristics of each vertex. Set them with
# %v%
add.health %v% "sex" <- cmty1$sex # we can do this because vertices and data
# are sorted in the same way
# A special case of vertex attributes is vertex names, which you set with
# network.vertex.names()
network.vertex.names(add.health) <- cmty1$ego_nid
# All of these methods can also be used to retrive network attributes
add.health %v% "sex"
# Third, edge attributes are characteristics of a pair of vertices. You can set
# them with %e%, but the indexing is more complicated, and I will not cover it
# here.
# These are useful for plotting and advanced statistical methods.
plot(add.health, vertex.col = "sex")
##### Managing lists #####
# data.frames are the R equivalent of a rectangular dataset in SAS, Stata, or
# SPSS. R's strength, however, is in its flexibility in the types of objects
# it can store. That's why it has become so popular for dealing with complex
# data sources, like networks or text. The data structure underlying all of
# these objects is a list.
# A list is exactly what it sounds like -- it is an ordered collection of
# objects. Those objects can be anything, including other lists!
my.list <- list(letters[1:10], dat, "all the things")
my.list
# We can refer to items in a list by number, using double brackets
my.list[[2]] # dat
# We can take more than one at a time using single brackets:
my.list[1:2] # letters[1:10] and dat
# Or we can name the elements in our list, and then refer to items by name
names(my.list) <- c("some_letters", "dat", "all_things")
my.list$all_things
my.list[["all_things"]] # same
# Why is this important? For starters, a data.frame is really just stored as
# a list of columns
dat$grade # prints just the column "grade"
str(dat) # view the data structure for dat
# This is helpful if we want to calculate values on that column
mean(dat$grade)
# It's also useful, because R can loop (i.e., perform the same action
# repeatedly) easily with lists.
# A typical loop looks something like this:
for (i in 1:3) {
print(my.list[[i]]) # prints each element of my.list, one at a time
}
# But that's speaking R with a strong C (or SAS) accent. The R way of doing
# this is to use the *apply family of functions. Each one of them takes a
# function, and applies it to each element of the list. They are named with a
# prefix that indicates what kind of output you'll get -- lapply returns a list,
# and sapply returns a simplified object.
# Let's replicate the example above.
lapply(my.list, print)
# Now let's take a more practical example -- what if I wanted to calculate the
# mean for each of the columns in dat?
lapply(dat, mean)
sapply(dat, mean) # note that this simplifies it into a vector
sapply(dat, mean, na.rm = T) # we can pass addition arguments to the function
dat.means <- sapply(dat, mean, na.rm = T) # save output
mean(dat$sdummy)
# Equivalent
dat.means <- c() # create empty vector for storage
for (i in 1:ncol(dat)) {
dat.means[i] <- mean(dat[[i]], na.rm = T)
}
# Notice the single brackets on the output. We're refering to a specific
# element of a vector by its index (this is called indexing). Indexing is
# another way to do recoding.
# For example, let's recode grade == 0 to NA (R's special character for missing)
dat[, "grade"] # prints the column for grade
dat$grade # equivalent
dat$grade == 0 # prints a logical (T/F) vector indicating when grade is 0
which(dat$grade == 0) # prints the index of which values are TRUE
dat[dat$grade == 0, "grade"] # prints just the elements of "grade" that are 0
dat$grade[dat$grade == 0] # equivalent
dat[dat$grade == 0, "grade"] <- NA # resets those values to NA
dat$grade[dat$grade == 0] <- NA # equivalent
# You would repeat this for each category. (This is cumbersome compared to
# dplyr!)
# The *apply functions are useful whenever you want to loop through several
# things. For example, here's how you would run several different model
# specifications for a linear model.
# One linear model looks like this:
lm(POP_BEH ~ IDG_SS, data = dat)
# (For the record, a logistic regression looks like this:)
glm(PSMOKES ~ IDG_SS, data = dat, family = binomial(link = "logit"))
# Let's specify a list of models:
models <- c( # Again, c stands for 'closure' -- it's a way of making a vector
POP_BEH ~ IDG_SS + grade,
POP_BEH ~ grade,
POP_BEH ~ sex
)
# Now, to run all the models, we use lm with lapply
(fits <- lapply(models, lm, data = dat))
# Print the summaries for both the models:
lapply(fits, summary)
# Ta-da! The object that we saved is a list of fitted model outputs. You could
# similarly run the model using different subsets of the data:
sex.models <- lapply(1:2, function(x) lm(POP_BEH ~ IDG_SS, data = dat,
subset = sex == x))
# The next section will explain how that last lapply statement worked.
##### Functions #####
# The other reason that R has become so popular is that it is very easy to build
# your own functions. Unlike SAS macros (or, to a lesser extent, Stata ado
# files), custom R functions are treated just like the built-in functions, and
# have a centralized distribution system (CRAN). This has made it very popular
# for statisticians, who use it to distribute software to do cutting edge
# analyses.
# You can, and sometimes should, build your own functions to make your code
# easier to read and maintain. The underlying goal is DRY: don't repeat
# yourself. If you find yourself copying and pasting code often, or if you
# have trouble parsing what a nested set of commands means, you should probably
# be writing a function.
# At their core, functions take an input, and produce an output. We call input
# the set of arguments that the function takes, and we call the output the
# return value of the function.
# Here's a simple function to calculate the mean manually:
newMean <- function(x) {
# Calculates the mean of a numeric vector
#
# Args:
# x: a numeric vector
#
# Returns:
# the arithmatic mean of x
return(sum(x) / length(x))
}
newMean(1:10)
mean(1:10) # same
# Because R is open-source, you can view the code for almost any function. For
# example, look at the code for lm:
lm
# Virtually all functions come with a help file, too. Access that with:
?lm # shows arguments, return values, etc.
# Of course, it's typically unwise to recreate existing functions, if for no
# other reason than that it's time consuming. Creating functions is useful
# because you can create custom output. For example, let's create a function
# that produces a PROC MEANS style summary of a numeric variable (mean, sd, min,
# max)
procMeans <- function(x) {
# Creates a data.frame with the mean, sd, min, and max of a numeric variable
#
# Args:
# x: a numeric vector
#
# Returns:
# a data.frame, with columns for the mean, sd, min, and max of x
return(data.frame(mean = mean(x), sd = sd(x), min = min(x), max = max(x)))
}
# Running this on one variable of dat:
procMeans(dat$IDG_SS)
# But, since a data.frame is just a list of columns, we can use lapply to loop
# through a bunch of columns at once:
lapply(dat[c("IDG_SS", "POP_BEH_SS", "PSMOKES", "grade")], procMeans)
# Not quite what we wanted -- let's pipe the result to the bind_rows function
# from dplyr, which squashes together the data.frames by rows
lapply(dat[c("IDG_SS", "POP_BEH_SS", "PSMOKES", "grade")], procMeans) %>%
bind_rows
# We will see that this method is useful when we are calculating a series of
# statistics on networks, or on individuals in networks.
