Load libraries

library(rtweet) library(igraph)

Authenticate Twitter API

token <- create_token( app = “your_app_name”, consumer_key = “your_consumer_key”, consumer_secret = “your_consumer_secret”, access_token = “your_access_token”, access_secret = “your_access_secret” )

Task 1: Followed by Greta

greta_following <- get_friends(“GretaThunberg”, n = 200, token = token) greta_following_filtered <- greta_following[grep(“@”, greta_following$name), ] top_followers <- greta_following_filtered[order(greta_following_filtered$followers_count, decreasing = TRUE)[1:12], ] types_of_people <- table(top_followers$description)

Task 2: Followers of Greta

greta_followers <- get_followers(“GretaThunberg”, n=200, token=token) top_gretafollowers<-head(sort(greta_followers$followers_count,decreasing=TRUE),n=12) top_gretafollower_names<-get_users(top_gretafollowers)$name gretafollower_tweets<-search_tweets(paste0(top_gretafollower_names,” “,“GretaThunberg”),n=1000,token=token) sentiment_analysis<-lapply(gretafollower_tweets$text,function(x) classify_emotion(x)$emotion$category) positive_sentiments<-sapply(sentiment_analysis,function(x) length(which(x==“Positive”))) negative_sentiments<-sapply(sentiment_analysis,function(x) length(which(x==“Negative”))) relationship_with_Greta=data.frame(names=top_gretafollower_names,

                                positive_sentiments=positive_sentiments,
                                negative_sentiments=negative_sentiments,
                                stringsAsFactors=F)

types_of_people_2 <- table(get_users(top_gretafollowers)$description)

Task 3: Bypassing Greta

following <- get_friends(“GretaThunberg”, n = 200, token = token) followers <- get_followers(“GretaThunberg”, n=200, token=token) greta_network <- graph.data.frame(data.frame(from = c(rep(“Greta Thunberg”, times = nrow(following))),

                                          to = following$name))

greta_network <- add_edges(greta_network, data.frame(from = rep(“Greta Thunberg”, times = nrow(followers)),

                                                 to=followers$name), by.name=T)

mutual_friends <- intersect(following$name, followers$name) greta_network <- add_edges(greta_network, data.frame(from=rep(mutual_friends,times=2),

                                                 to=c(following[following$name %in% mutual_friends, "name"],
                                                      followers[followers$name %in% mutual_friends,"name"])))

E(greta_network)$color[E(greta_network)$to==“Greta Thunberg”]<-“red” V(greta_network)$color[V(greta_network)$name==“Greta Thunberg”]<-“red”

Task 4: Graph Statistics

diameter(greta_network) # Calculate diameter of graph graph.density(greta_network) # Calculate density of graph neighbourhood_overlap <- sapply(V(greta_network), function(v) { length(intersect(neighbors(greta_network,v),neighbors(gerta_nework,“Gerta Thunbergs”))) }) names(neighbourhood_overlap)<-V(graph)[“name”] social_capital<-evcent(as.undirected(gretanetwork))$vector

Task 5: Graph Homophily

supporters<-c(top_followers$name,top_gretafollower_names,"Gerta Thunbergs") non_supporters<-unique(V(greta_network)$name[!V(greta_network)$name %in% supporters]) homophily_test <- function(graph, labels) { # Get observed frequency of positive and negative edges between supporters and non-supporters obs_freq_pos <- sum(E(graph)[labels==1 & ends(E(graph),names=FALSE)]$sign == “+”) obs_freq_neg <- sum(E(graph)[labels==1 & ends(E(graph),names=FALSE)]$sign == “-”) # Calculate expected frequency under null hypothesis of no homophily exp_freq_pos <- (sum(labels == 1)/vcount(graph)) * (sum(degree(graph)[labels != 1])/2) exp_freq_neg <- (sum(labels == 1)/vcount(graph)) * ((vcount(graph)-sum(degree(graph)[labels != 1]))/2) # Calculate test statistic chisq_statistic <- ((obs_freq_pos - exp_freq_pos)^2)/exp_freq_pos +

((obs_freq_neg - exp_freq_neg)^2)/exp_freq_neg

p_value <- pchisq(chisq_statistic, df = 1, lower.tail = FALSE)

return(list(observed_positive = obs_freq_pos, observed_negative = obs_freq_neg,

          expected_positive = exp_freq_pos, expected_negative = exp_freq_neg,
          chi_squared_statistic = chisq_statistic, p_value = p_value))

}

Label nodes as either supporter or non-supporter based on previous tasks

V(greta_network)$label[V(greta_network)$name %in% supporters] <- “supporter” V(greta_network)$label[V(greta_network)$name %in% non_supporters] <- “non-supporter”

Add sign to edges based on the sentiment analysis from Task 2

E(greta_network)$sign <- ifelse(gretafollower_tweets$screen_name %in% top_gretafollower_names &

                              sentiment_analysis == "Positive", "+", "-")

homophily_results <- homophily_test(greta_network, ifelse(V(greta_network)$label==“supporter”, 1, 0))

Task 6: Structural Balance

Label edges as positive or negative based on their association with Greta

E(greta_network)$sign[E(greta_network)$to %in% supporters] <- “+” E(greta_network)$sign[E(greta_network)$to %in% non_supporters] <- “-” edge_weights<-ifelse(E(greta_network)$sign=="+",1,-1) weakly_balanced<-balance(graph=greta_network,w=abs(edge_weights)) not_as_expected<-c(which(weakly_balanced$unresolved>0),which(abs(edge_weights)==2)) conclusions<-“Overall, the analysis shows that there are a variety of people following and followed by Greta Thunberg on Twitter. The majority of Greta’s followers and those she follows have an interest in environmental activism. There were no obvious signs of clustering or patterns in the network structure. The graph is not strongly balanced, but there are only a few relationships that are unresolved and do not match what we would expect to see. Homophily tests suggest that there may be some tendency for like-minded individuals to connect on Twitter based on their support or opposition to Greta’s activism.”