Analyses of resolutions.R

library(ggplot2)

positions = c("Travel and Holidays", "Finances", "Learning and Career", "Mental Wellbeing",
              "Relationships", "Physical Health")

# What are the most popular resolutions?
p1 <- ggplot(twitter_df[twitter_df$Resolution.type != "",], aes(x = Resolution.type, fill = Resolution.type)) +
        geom_bar() +
        coord_flip() +
        ggtitle("Number of tweets by resolution type") +
        scale_x_discrete(limits = positions, name = "") +
        scale_y_continuous(name = "Number of tweets") +
        scale_fill_brewer(palette = "Paired") +
        theme_bw() +
        theme(plot.title = element_text(size = 12, family = "Tahoma", face = "bold"),
              text = element_text(size = 10, family = "Tahoma")) +
        labs(fill = "Resolution") + 
        geom_text(stat = "count", aes(label = ..count..),
                  family = "Tahoma", size = 3, hjust = 1)
p1

# How do people feel about their resolutions?
p2 <- ggplot(data = twitter[twitter$Resolution.type != "" & twitter$Compound != 0,], 
             aes(x = Compound, colour = Resolution.type)) +
        geom_density(position="identity", fill = NA, size = 0.8) +
        ggtitle("Density plots of overall sentiment by resolution type") +
        scale_x_continuous(name = "Compound sentiment score") +
        scale_y_continuous(name = "Density") +
        scale_colour_brewer(palette = "Paired") +
        theme_bw() +
              theme(plot.title = element_text(size = 12, family = "Tahoma", face = "bold"),
                    text = element_text(size = 10, family = "Tahoma"),
                    legend.position="none") +
        facet_wrap( ~ Resolution.type, ncol=2)
p2

# Get median compound resolutions
library(knitr)

kable(aggregate(Compound ~ Resolution.type, 
                data = twitter[twitter$Compound != 0 & twitter$Resolution.type != "",], 
                FUN = median), digits = 3, col.names = c("Resolution", "Sentiment Score"))
                
# How do people’s friends feel about their resolutions?
library(plyr)

agg_twitter <- ddply(twitter[twitter$Resolution.type != "" & twitter$Compound != 0,], 
                     "Resolution.type", summarise,
                     mdn.compound = median(Compound),
                     num.resolutions = length(Resolution.type),
                     prop.favourite = sum(Five.or.more.favourites) / length(Resolution.type))

p3 <- ggplot(data = agg_twitter, aes(x = mdn.compound, y = prop.favourite, 
                               size = num.resolutions,
                               colour = Resolution.type)) +
        geom_point() +
        scale_size_area(max_size = 25) +
        ggtitle("Weighted scatterplot of tweet sentiment against number of favourites") +
        scale_x_continuous(limits = c(0.1, 0.6), name = "Compound sentiment score (median)") +
        scale_y_continuous(limits = c(0.14, 0.23), name = "Received five or more favourites (%)") +
        scale_colour_brewer(palette = "Paired") +
        labs(colour = "Resolution", size = "Number of tweets") +
        theme_bw() +
              theme(plot.title = element_text(size = 12, family = "Tahoma", face = "bold"),
                    text = element_text(size = 10, family = "Tahoma"))
p3