Visualization of statistical hypothesis testing by Ruslana Moskotina (Refactoring)

Alternative.R

# Include necessary libraries
library(foreign)
library(igraph)
library(ggraph)

# Download data here 
# https://www.europeansocialsurvey.org/download.html?file=ESS6UA&c=UA&y=2012

dataESS <- read.spss('ESS6UA.sav', to.data.frame = T)

# Check if person is religious
rlgblg <- dataESS$rlgblg
levels(rlgblg) <- c('Так', 'Ні')

# Recode happiness variable
happy <- cut(as.numeric(dataESS$happy), c(0, 4, 7, 11), c('Ні','Нрм','Так'))

# Recode gender
gndr <- dataESS$gndr
levels(gndr) <- c('Ч', 'Ж')

# Tendency to help
help <- as.numeric(dataESS$prhlppl) - 1

# Create groups
group <- paste(rlgblg, happy, gndr, sep = ".")
group[grepl('NA', group)] <- NA
level <- names(sort(tapply(help, group, mean, na.rm = T), decreasing = T))
group <- factor(group, levels = level)

# Kruskal Test
kruskal.test(help ~ group)

# Pairwise testing
p_val <- pairwise.wilcox.test(help, group, 
                              p.adjust.method = 'BH', 
                              exact = F) [['p.value']]

# Create a directed distance matrix (to make a directed graph)
dim = nrow(p_val) + 1
graph_m <- matrix(rep(0, dim^2), nrow = dim)
graph_m[upper.tri(graph_m)] <- t(p_val)[!lower.tri(p_val)]
colnames(graph_m) <- c(colnames(p_val)[1], rownames(p_val))

l <- layout_(net, as_star())

layout_with_kk(net, weights = E(net)$weight)
layout_with_fr(net, weights = E(net)$weight)

# Create a network
net <- graph_from_adjacency_matrix(graph_m, weighted = T)
ggraph(net, layout = layout_with_fr(net, weights = E(net)$weight)) +
  geom_edge_link(aes(alpha = weight, width=weight)) +
  scale_edge_width(range = c(0.01, 0.5)) +
  geom_node_point(size = 4, color = color) +
  geom_node_text(aes(label = level), 
                 size = 3.2, 
                 fontface = 'bold',
                 ) +
  xlab('') +
  ylab('') +
  theme(legend.position = 'none')

Moskotina.R

# Include necessary libraries
library(foreign)
library(igraph)
library(ggraph)

# Download data here 
# https://www.europeansocialsurvey.org/download.html?file=ESS6UA&c=UA&y=2012

dataESS <- read.spss('ESS6UA.sav', to.data.frame = T)

# Check if person is religious
rlgblg <- dataESS$rlgblg
levels(rlgblg) <- c('Так', 'Ні')

# Recode happiness variable
happy <- cut(as.numeric(dataESS$happy), c(0, 4, 7, 11), c('Ні','Нрм','Так'))

# Recode gender
gndr <- dataESS$gndr
levels(gndr) <- c('Ч', 'Ж')

# Tendency to help
help <- as.numeric(dataESS$prhlppl) - 1

# Create groups
group <- paste(rlgblg, happy, gndr, sep = ".")
group[grepl('NA', group)] <- NA
level <- names(sort(tapply(help, group, mean, na.rm = T), decreasing = T))
group <- factor(group, levels = level)

# Kruskal Test
kruskal.test(help ~ group)

# Pairwise testing
p_val <- pairwise.wilcox.test(help, group, 
                              p.adjust.method = 'BH', 
                              exact = F) [['p.value']]

# Save only significant differences
p_val <- apply(p_val, 1:2, function(x) { sum(x <= 0.05) })

# Create a directed distance matrix (to make a directed graph)
dim = nrow(p_val) + 1
graph_m <- matrix(rep(0, dim^2), nrow = dim)
graph_m[upper.tri(graph_m)] <- t(p_val)[!lower.tri(p_val)]
colnames(graph_m) <- c(colnames(p_val)[1], rownames(p_val))

# Create a network
net <- graph_from_adjacency_matrix(graph_m)

# Create a symmetric matrix for MDS
mds_m <- matrix(rep(0, dim^2), nrow = dim)
mds_m[lower.tri(mds_m)] <- p_val[!upper.tri(p_val)]
mds_m[upper.tri(mds_m)] <- t(mds_m)[upper.tri(mds_m)]


# Create MDS layout
l <- layout_with_mds(net, dist = mds_m)
l[,c(1, 2)] <- l[,c(2, 1)]

# Get degrees
in_degree <- degree(net, mode = 'in')
out_degree <- degree(net, mode = 'out')

# Define graph colors
color <- rep('gold', 12)
color[in_degree == 0] <- 'deepskyblue2'
color[out_degree == 0] <- 'coral2'

# Grouping of edges
edges <- as_edgelist(net)
col_from <- rep('gold', gsize(net))
col_from[is.na(match(edges[, 1], edges[, 2]))] <- 'deepskyblue2'

col_to <- rep('gold', gsize(net))
col_to[is.na(match(edges[, 2], edges[, 1]))] <- 'coral2'
edge_gr <- rep('03', gsize(net))
edge_gr[col_from == 'deepskyblue2' & col_to == 'gold'] <- '01'
edge_gr[col_from == 'deepskyblue2' & col_to == 'coral2'] <- '02'

# Plot finally
ggraph(net, layout = l) +
  theme_gray() +
  xlim(-0.75, 0.5) +
  ylim(-0.5, 0.75) +
  geom_edge_arc(aes(colour = edge_gr),
                strength = c(rep(0, 24), 0.59, 0, 0, 0),
                show.legend = F,
                arrow = arrow(length = unit(0.1, 'inches')),
                end_cap = circle(3.5, 'mm'),
                start_cap = circle(3.5, 'mm')) +
  scale_edge_colour_manual(values = c('green3', 'purple3', 'orange')) +
  geom_node_point(size = 4,
                  color = color) +
  geom_node_text(aes(label = level), 
                 size = 3.2, 
                 fontface = 'bold',
                 position = position_nudge(x = c(0, 0, 0.03, 0.05, 0, 0.01, -0.02, -0.08, 0.04, -0.04, 0, 0),
                                           y = c(rep(0.05, 4), 0.01, -0.04, -0.04, 0, rep(-0.05, 4)))) +
  xlab('') +
  ylab('Менша схильність допомагати - Більша схильність допомагати')