Quick and dirty analysis of iPhone Sales and Google Trends

google-trends.R

library(ggplot2)
require(reshape2)
library(dplyr)
library(lubridate)
library(scales)
options(scipen=999)
options(stringsAsFactors = FALSE)

sales <- read.csv("quarterly-iphone-sales.csv")
ggplot(sales, aes(x=quarter, y=sales)) + geom_point()

queries <- read.csv("iphone-ipad.csv")
queries$date <- mdy(as.character(queries$date))
queries$quarter <- paste(year(queries$date), quarters(queries$date))

queries.agg <- queries %.% group_by(quarter) %.% summarise(queries=mean(iphone))


queries.agg.join <- left_join(queries.agg, sales)

sales <- inner_join(sales, queries.agg)
sales$norm_sales <- scale(sales$sales)
sales$norm_queries <- scale(sales$queries)

sales.pt <- melt(sales[, c("quarter", "norm_sales", "norm_queries")], "quarter")
plot <- ggplot(filter(sales.pt, quarter > "2008 Q1"), aes(x=gsub(" ", "\n", quarter), y=value, color=variable)) + geom_point() + ggtitle("Normalized Apple iPhone Sales and Google Search Query Index") + xlab("Quarter") + ylab("") + scale_color_discrete(breaks=c("norm_sales", "norm_queries"), labels=c("Sales", "Queries"), name="Normalized\nVariable") + theme(legend.position=c(.9,.2),legend.key = element_rect(fill="white"), legend.background = element_rect(fill="white"), axis.title.y=element_blank())
ggsave("norm-sales-queries.png", plot, width=640 / 72, height=400 / 72, dpi=72)
model <- glm(sales ~ queries, data=sales)
queries.agg.join$pred_sales <- predict(model, queries.agg.join)

plot <- ggplot(filter(queries.agg.join, quarter > "2008 Q1", quarter < "2014 Q1"), aes(x=gsub(" ", "\n", quarter), y=sales)) + geom_bar(stat="identity") + geom_bar(stat="identity", data=filter(queries.agg.join, quarter == "2013 Q4"), aes(x=gsub(" ", "\n", quarter), y=pred_sales), fill="red") + ggtitle("Apple iPhone Sales (in Millions of Units)") + xlab("Quarter") + theme(axis.title.y=element_blank(), axis.title.x=element_blank())
ggsave("pred-sales.png", plot, width=640 / 72, height=400 / 72, dpi=72)