Assignment 3 - Social Network Analysis
+Part I
+Start by installing the “igraph” package. Once you have installed igraph, load the package.
+Now upload the data file “comment-data.csv” as a data frame called “D1”. Each row represents a comment from one student to another so the first line shows that student “28” commented on the comment of student “21”. It also shows the gender of both students and the students’ main elective field of study (“major”").
+D1 <- read.csv("comment-data.csv", header = TRUE)Before you proceed, you will need to change the data type of the student id variable. Since it is a number R will automatically think it is an integer and code it as such (look at the list of variables by clicking on the data frame arrow in the Data pane. Here you will see the letters “int”" next to the stid variable, that stands for integer). However, in this case we are treating the variable as a category, there is no numeric meaning in the variable. So we need to change the format to be a category, what R calls a “factor”. We can do this with the following code:
+D1$comment.to <- as.factor(D1$comment.to)
+D1$comment.from <- as.factor(D1$comment.from)igraph requires data to be in a particular structure. There are several structures that it can use but we will be using a combination of an “edge list” and a “vertex list” in this assignment. As you might imagine the edge list contains a list of all the relationships between students and any characteristics of those edges that we might be interested in. There are two essential variables in the edge list a “from” variable and a “to” variable that descibe the relationships between vertices. While the vertex list contains all the characteristics of those vertices, in our case gender and major.
+So let’s convert our data into an edge list!
+First we will isolate the variables that are of interest: comment.from and comment.to
+library(dplyr)##
+## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
+##
+## filter, lag## The following objects are masked from 'package:base':
+##
+## intersect, setdiff, setequal, unionD2 <- select(D1, comment.to, comment.from) #select() chooses the columnsSince our data represnts every time a student makes a comment there are multiple rows when the same student comments more than once on another student’s video. We want to collapse these into a single row, with a variable that shows how many times a student-student pair appears.
+EDGE <- count(D2, comment.to, comment.from)
+
+names(EDGE) <- c("from", "to", "count")EDGE is your edge list. Now we need to make the vertex list, a list of all the students and their characteristics in our network. Because there are some students who only recieve comments and do not give any we will need to combine the comment.from and comment.to variables to produce a complete list.
+#First we will separate the commenters from our commentees
+V.FROM <- select(D1, comment.from, from.gender, from.major)
+
+#Now we will separate the commentees from our commenters
+V.TO <- select(D1, comment.to, to.gender, to.major)
+
+#Make sure that the from and to data frames have the same variables names
+names(V.FROM) <- c("id", "gender.from", "major.from")
+names(V.TO) <- c("id", "gender.to", "major.to")
+
+#Make sure that the id variable in both dataframes has the same number of levels
+lvls <- sort(union(levels(V.FROM$id), levels(V.TO$id)))
+
+VERTEX <- full_join(mutate(V.FROM, id=factor(id, levels=lvls)),
+ mutate(V.TO, id=factor(id, levels=lvls)), by = "id")
+
+#Fill in missing gender and major values - ifelse() will convert factors to numerical values so convert to character
+VERTEX$gender.from <- ifelse(is.na(VERTEX$gender.from) == TRUE, as.factor(as.character(VERTEX$gender.to)), as.factor(as.character(VERTEX$gender.from)))
+
+VERTEX$major.from <- ifelse(is.na(VERTEX$major.from) == TRUE, as.factor(as.character(VERTEX$major.to)), as.factor(as.character(VERTEX$major.from)))
+
+#Remove redundant gender and major variables
+VERTEX <- select(VERTEX, id, gender.from, major.from)
+
+#rename variables
+names(VERTEX) <- c("id", "gender", "major")
+
+#Remove all the repeats so that we just have a list of each student and their characteristics
+VERTEX <- unique(VERTEX)Now we have both a Vertex and Edge list it is time to plot our graph!
+#Load the igraph package
+
+library(igraph)##
+## Attaching package: 'igraph'## The following objects are masked from 'package:dplyr':
+##
+## as_data_frame, groups, union## The following objects are masked from 'package:stats':
+##
+## decompose, spectrum## The following object is masked from 'package:base':
+##
+## union#First we will make an object that contains the graph information using our two dataframes EDGE and VERTEX. Notice that we have made "directed = TRUE" - our graph is directed since comments are being given from one student to another.
+
+g <- graph.data.frame(EDGE, directed=TRUE, vertices=VERTEX)
+
+#Now we can plot our graph using the force directed graphing technique - our old friend Fruchertman-Reingold!
+
+plot(g,layout=layout.fruchterman.reingold)
#There are many ways to change the attributes of the graph to represent different characteristics of the newtork. For example, we can color the nodes according to gender.
+
+plot(g,layout=layout.fruchterman.reingold, vertex.color=VERTEX$gender)
#We can change the thickness of the edge according to the number of times a particular student has sent another student a comment.
+
+plot(g,layout=layout.fruchterman.reingold, vertex.color=VERTEX$gender, edge.width=EDGE$count)
Part II
+In Part II your task is to look up in the igraph documentation and modify the graph above so that:
+-
+
- Ensure that sizing allows for an unobstructed view of the network features (For example, the arrow size is smaller) +
- The vertices are colored according to major +
- The vertices are sized according to the number of comments they have recieved +
plot(g, edge.arrow.size=.4, edge.curved=.1)
colrs <- c("gray50", "pink", "gold")
+
+V(g)$color <- colrs[V(g)$major]
+
+E(g)$arrow.size <- .2
+
+V(g)$size <- V(g)$comment.to*0.7## Warning in length(vattrs[[name]]) <- vc: length of NULL cannot be changedE(g)$width <- 1+E(g)$comment.from/12## Warning in length(eattrs[[name]]) <- ec: length of NULL cannot be changedplot(g)
plot(g, vertex.shape="none", vertex.label=V(g)$first.name,
+
+ vertex.label.font=2, vertex.label.color="gray40",
+
+ vertex.label.cex=.7, edge.color="gray85")
Part III
+Now practice with data from our class. This data is real class data directly exported from Qualtrics and you will need to wrangle it into shape before you can work with it. Import it into R as a data frame and look at it carefully to identify problems.
+Please create a person-network with the data set hudk4050-classes.csv. To create this network you will need to create a person-class matrix using the tidyr functions and then create a person-person matrix using t(). You will then need to plot a matrix rather than a to/from data frame using igraph.
library(tidyr) library(dplyr) library(stringr) library(igraph)
+#Input Data
+C1 <- read.csv(“hudk4050-classes.csv”, stringsAsFactors = FALSE, header = TRUE)
+C2 <- C1 View(C2)
+Data Tidying
+Make header first row
+colnames(C2) <- C2[1,]
+Remove unwanted column
+C2 <- slice(C2, 3:49)
+Remove last column
+C2 <- select(C2, 1:8)
+Merger name columns
+C2 <- unite(C2, “name”, ‘First Name’,‘Last Name’, sep =" ")
+Remove unpredictable characters from names
+C2\(name <- str_replace(C2\)name, “`”,"")
+Make all names capitalized first letters only
+C2\(name <- str_to_title(C2\)name)
+Make all class letters capitals
+C2 <- C2 %>% mutate_at(2:7, list(toupper)
+Remove whitespace between letters and numbers in class
+C2 <- C2 %>% mutate_at(2:7, str_replace_all, " “,”")
+Data restructing
+#Create a new variable with two variables, student and class C3 <- C2 %>% gather(label1, class, 2:7, na.rm = TRUE, convert =FALSE) %>% select(name, class)
+#Create a new variable containing 1s that will become counts C3$count <- 1
+#Remove blank classes C3 <- filter(C3, class != "")
+#Remove duplicates (Danny!) C3 <-unique(C3)
+#Spread 1s across classes to create a student x class dataframe C3 <- spread(C3, class, count)
+#Make row names student names rownames(C3) <- C3$name
+#Remove names column AND HUDK 4050 C3 <- select(C3, -name, -HUDK4050) #Shortest: C3[is.na(C3)] <- 0
+#Matrix Operations #Convert to Matrix C4 <- as.matrix(C3) C4 <- C4 %*% t(C4)
+#Graphing g <- graph.adjacency(C4, mode=“undirected”, diag =FALSE) plot(g, layout=layout.fruchterman.reingold, vertex.size=4, vertex.label.cex=0.8,vertex.label.color=“black”,vertex,color=“gainsboro”)
+#Centrality
+sort(degree(g), decreasing= TRUE)## 7 16 15 27 21 24 4 23 29 18 28 6 17 20 26 25 11 5 2 19 12 3 10 1 22 8
+## 8 7 6 6 6 6 5 5 5 5 4 4 4 4 4 4 3 3 3 3 3 2 2 2 2 2
+## 14 13 9
+## 2 1 1sort(betweenness(g),decreasing= TRUE)## 7 16 23 20 24 26 27
+## 127.666667 119.916667 80.333333 79.166667 71.833333 65.833333 61.333333
+## 18 15 21 6 28 10 12
+## 54.000000 47.583333 27.916667 17.000000 16.000000 15.833333 12.000000
+## 29 4 19 3 11 17 5
+## 6.583333 6.500000 4.500000 0.000000 0.000000 0.000000 0.000000
+## 2 13 1 9 22 25 8
+## 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000
+## 14
+## 0.000000Yifei has the highest betweeness in centrality measurement because she gets the score of 260.614, which is a lot higher than other students. This implies that she could be the connect person for our class. And there is a couple of students (i.e.Guoliang, Hangshi, etc) has the high degree.
+