-
Notifications
You must be signed in to change notification settings - Fork 3
/
analysis.R
188 lines (156 loc) · 6.72 KB
/
analysis.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
library(ggplot2)
library(scales)
library(dplyr)
library(reshape2)
library(readr)
library(rpart)
library(rpart.plot)
library(grid)
source("helpers.R")
stats = read_csv("learnedleague_category_stats.csv")
correct_rates = dcast(stats, anon_id + gender + overall_correct_pct ~ category, value.var = "ratio")
outperformance = dcast(stats, anon_id + gender + overall_correct_pct ~ category, value.var = "ratio_relative_to_overall")
categories = sort(unique(stats$category))
display_names = c(
amer_hist = "American History",
art = "Art",
bus_econ = "Business/Economics",
class_music = "Classical Music",
curr_events = "Current Events",
film = "Film",
food_drink = "Food/Drink",
games_sport = "Games/Sport",
geography = "Geography",
language = "Language",
lifestyle = "Lifestyle",
literature = "Literature",
math = "Math",
pop_music = "Pop Music",
science = "Science",
television = "Television",
theatre = "Theatre",
world_hist = "World History"
)
# calculate correlations
# assumes input_data is a data frame that has a column for each category
calculate_correlation_pairs = function(input_data) {
correlation_matrix = cor(input_data[, categories])
rownames(correlation_matrix) = display_names[rownames(correlation_matrix)]
colnames(correlation_matrix) = display_names[colnames(correlation_matrix)]
correlation_pairs = melt(correlation_matrix)
names(correlation_pairs) = c("category_1", "category_2", "rho")
correlation_pairs = filter(correlation_pairs, as.character(category_1) < as.character(category_2))
correlation_pairs = correlation_pairs[rev(order(correlation_pairs$rho)), ]
correlation_pairs
}
correct_rates_correlations = calculate_correlation_pairs(correct_rates)
correct_rates_correlations$type = "absolute correct rate"
outperformance_correlations = calculate_correlation_pairs(outperformance)
outperformance_correlations$type = "relative to individual overall average"
write_csv(rbind(correct_rates_correlations, outperformance_correlations), "correlation_pairs.csv")
# scatterplots of category pairs
plot_categories = function(x, y) {
p = ggplot(data = correct_rates, aes_string(x = x, y = y)) +
geom_point(color = "#900000", alpha = 0.15, size = 4.5) +
geom_smooth(method = "lm", se = FALSE) +
scale_x_continuous(paste0("\n", display_names[x], " questions correct %"), labels = percent) +
scale_y_continuous(paste0(display_names[y], " questions correct %\n"), labels = percent) +
title_with_subtitle(paste0(display_names[y], " vs. ", display_names[x]),
paste0("Based on trivia results from ", nrow(correct_rates), " LearnedLeague players")) +
theme_tws(base_size = 18) +
theme(plot.margin = unit(c(1, 1, 1.25, 0.5), "lines"))
p
}
apply(expand.grid(categories, categories), 1, function(row) {
x = row[1]
y = row[2]
filename = paste0("graphs/", x, "__", y, ".png")
png(filename = filename, width = 480, height = 480)
print(plot_categories(x, y))
add_credits(fontsize = 10)
dev.off()
})
# barplots for categories
category_barplot = function(category) {
data = filter(correct_rates_correlations, category_1 == category | category_2 == category)
other = data$category_1
other[which(other == category)] = data$category_2[which(other == category)]
data$other = as.character(other)
data = data[rev(order(data$rho)), ]
data$other = factor(data$other, levels = data$other)
p = ggplot(data = data, aes(x = other, y = rho)) +
geom_bar(stat = "identity", fill = "#120060") +
scale_y_continuous("correlation\n", lim = c(0, 1)) +
scale_x_discrete("") +
labs(title = paste0(category, " correlation to other categories\n")) +
theme_tws(base_size = 18) +
theme(axis.text.x = element_text(angle = -90, vjust = 0.5, hjust = 0.05),
legend.position = "none")
p
}
for(category in categories) {
display_name = display_names[category]
png(filename=paste0("graphs/", category, "_barplot.png"), width=640, height=480)
print(category_barplot(display_name))
add_credits(fontsize = 10)
dev.off()
}
# gender classification tree
set.seed(1738)
model_data = filter(correct_rates, gender %in% c("Male", "Female"))[, c("gender", categories)]
names(model_data) = c("Gender", display_names)
rpart_model = rpart(Gender ~ ., data = model_data, parms = list(prior = c(0.5, 0.5)))
printcp(rpart_model)
plotcp(rpart_model)
# prune tree based on CP table
pruned_model = prune(rpart_model, cp = 0.014)
table(model_data$Gender, predict(rpart_model, type = "class"))
table(model_data$Gender, predict(pruned_model, type = "class"))
pruned_model$variable.importance
png(filename = "graphs/gender_decision_tree.png", width = 640, height = 1080)
par(bg = "#f4f4f4")
prp(rpart_model,
type = 3, extra = 8,
main = "Gender classification tree based on category performance\nin LearnedLeague trivia competition",
clip.right.labs = FALSE, branch = 1, varlen = 0,
cex = 1.1, cex.main = 1.7,
box.col = c("pink", "skyblue")[rpart_model$frame$yval])
add_credits(fontsize = 14)
dev.off()
png(filename = "graphs/pruned_gender_decision_tree.png", width = 640, height = 720)
par(bg = "#f4f4f4")
prp(pruned_model,
type = 3, extra = 8,
main = "Gender classification tree based on category performance\nin LearnedLeague trivia competition",
clip.right.labs = FALSE, branch = 1, varlen = 0,
cex = 1.1, cex.main = 1.7,
box.col = c("pink", "skyblue")[pruned_model$frame$yval])
add_credits(fontsize = 14)
dev.off()
# ranking the categories
gender_stats = summarize(
group_by(filter(stats, gender %in% c("Male", "Female")),
gender, category),
outperformance = sum(ratio_relative_to_overall * total) / sum(total)
)
gender_stats$category = display_names[gender_stats$category]
gender_diffs = summarize(
group_by(gender_stats, category),
outperformance_diff = sum(outperformance * (gender == "Male")) - sum(outperformance * (gender == "Female"))
)
gender_diffs = gender_diffs[order(gender_diffs$outperformance_diff), ]
gender_diffs$category = factor(gender_diffs$category, levels = gender_diffs$category)
png(filename = "graphs/category_preferences.png", width = 640, height = 640)
ggplot(data = gender_diffs, aes(x = category, y = outperformance_diff, fill = factor(sign(outperformance_diff)))) +
geom_bar(stat = "identity", position = "dodge") +
scale_y_continuous("(male - female) preference\n", labels = percent) +
scale_x_discrete("") +
scale_fill_manual("", values = c("pink", "skyblue")) +
labs(title = "LearnedLeague trivia categories ranked\nby Male/Female preference\n") +
theme_tws(base_size = 18) +
theme(axis.text.x = element_text(angle = -90, vjust = 0.5, hjust = 0.05),
legend.position = "none")
add_credits(fontsize = 14)
dev.off()
# logistic regression
summary(glm(factor(Gender) ~ ., data = model_data, family = binomial(link = "logit")))