setwd("/Users/lotze/workspace/sfjourney_analysis")
require('RColorBrewer')

registrations = read.csv("cleaned_registrations.csv", stringsAsFactors=FALSE)
registrations = registrations[order(registrations$signup_timestamp), ]

##### Section 1 #####

registrations$num_reg = (1:nrow(registrations))
registrations$num_attended = cumsum(registrations$attended)
registrations$pct_attended = registrations$num_attended/registrations$num_reg

attended <- registrations[registrations$attended,]
attended$num_reg = (1:nrow(attended))

# histogram of signup ages
svg("age_histogram.svg", width=8, height=6)
hist(registrations$age, xlim=c(15,44), breaks=seq(-0.5,100.5), xlab="age", main="Histogram of Registrations by Age")
dev.off()
summary(registrations$age)
quantile(registrations$age, prob=c(0.01, 0.025, 0.05, 0.95, 0.975, 0.99))

# unused heatmap of gender/age
ages = seq(0, max(registrations$age), by=1)
registrations$gender = "undetermined"
registrations$gender[registrations$male > 0.5] = "M"
registrations$gender[registrations$male <= 0.5] = "F"
by_age = data.frame(age=ages, M=rep(0, length(ages)), F=rep(0, length(ages)))
M = table(registrations$age[registrations$gender=="M"])
by_age$M[match(names(M), by_age$age)] = M
F = table(registrations$age[registrations$gender=="F"])
by_age$F[match(names(F), by_age$age)] = F
by_age$total = by_age$F + by_age$M
rownames = by_age$age
by_age = by_age[, c("age", "M", "F", "total")]
heatmap(as.matrix(by_age[, c("M", "F")]/sum(by_age$total)), Rowv=NA, Colv=NA, scale="none")

# barplot of gender/age
svg("age_gender_barplot.svg", width=8, height=6)
main_ages = by_age[by_age$age >= 14 & by_age$age <=50,]
barplot(t(as.matrix(main_ages[, c("M", "F")])), names.arg=main_ages$age, beside=TRUE, legend=TRUE, xlab="age", ylab="registrations", main="Registrations by Age and Gender")
dev.off()

# gender comparison: about 1.5 (3 men to every 2 women)
num_males = sum(registrations$male > 0.5, na.rm=TRUE)
num_females = sum(registrations$male < 0.5, na.rm=TRUE)
num_females / (num_males + num_females)

svg("why_is_this_a_plot.svg", width=8, height=6)
barplot(as.matrix(c(num_males, num_females)), beside=TRUE, legend = c("Male", "Female"), ylab="registered", main="60% Male, 40% Female")
dev.off()

# interestingly, there does seem to be a slight age difference, with males skewing older
quantile(registrations$age[registrations$male > 0.5], prob=c(0.01, 0.025, 0.25, 0.5, 0.75, 0.975, 0.99), na.rm=TRUE)
quantile(registrations$age[registrations$male < 0.5], prob=c(0.01, 0.025, 0.25, 0.5, 0.75, 0.975, 0.99), na.rm=TRUE)
hist(registrations$age[registrations$male > 0.5], xlim=c(15,44), breaks=100)
hist(registrations$age[registrations$male < 0.5], xlim=c(15,44), breaks=100)

# interesting note: breakdown of registration methods
for (method in c("facebook", "google", "twitter")) {
    registrations[, sprintf("is_%s", method)] = grepl(method, registrations$services)
    print(sprintf("%s: %0.2f%%", method, 100*sum(registrations[, sprintf("is_%s", method)])/nrow(registrations)))
}
summary(registrations[,c("is_facebook", "is_google", "is_twitter")])
# wow! more people authenticated with facebook than google!
svg("authentication_bars.svg", width=8, height=6)
reg_percents = 100*sapply(registrations[,c("is_facebook", "is_google", "is_twitter")], sum)/nrow(registrations)
barplot(reg_percents, beside=TRUE, ylab="percent registrations using method", names.arg=c("facebook", "google", "twitter"), main="Authentication method Frequency")
dev.off()

# and breakdown of email domains
domain_table = table(registrations$domain)
domain_table = rev(sort(domain_table))
domain_names = names(domain_table)
domain_table = 100*domain_table/sum(domain_table)
names(domain_table) = domain_names
domain_table[1:30]
# yep; gmail is kind of email; yahoo and hotmail still alive; a few local universities
svg("email_bars.svg", width=8, height=6)
barplot(domain_table[1:6], ylab="percent registrations", main="Email Domain Frequency")
dev.off()

# and breakdown of area codes
area_code_table = rev(sort(table(registrations$area_code)))
area_code_table[1:20]
# 415 and 510, sure -- but a lot of diversity (maybe not surprisingly)
registrations$area_code_factor = registrations$area_code
registrations$area_code_factor[!(registrations$area_code %in% names(area_code_table[1:10]))] = "small"
registrations$area_code_factor = as.factor(registrations$area_code_factor)