Script to plot estimated twitter sizes by month, day-of-week, hour-of-day

esttimeline.csv

plotVols.r

#!/usr/bin/env Rscript

library(ggplot2)
library(gridExtra)
library(mgcv)

bytePerTweet <- 397.

#read data file with headers
X <-read.delim("./esttimeline.csv", sep=",", header=TRUE)

# make dates with number representation for plotting
X$dates <- as.POSIXct(X$date)
# make dates: dow factor  sunday = 0
X$dow <- as.factor(as.POSIXlt(X$date)$wday)
# make dates: hours factor
X$hour <- as.factor(as.POSIXlt(X$date)$hour)
#
X$approxTweets <- X$filesize/bytePerTweet
#
X$norm <- X$approxTweets - mean(X$approxTweets)
X$norm <- 1. + X$norm/mean(X$approxTweets)
X$fact <- 24./(X$norm * bytePerTweet)

mean(X$approxTweets)
24*mean(X$approxTweets)

# see some data
X[1:10,]

# stats
summary(X)
aggregate(data=X, cbind(links, mentions, filesize, approxTweets, norm, fact) ~ hour, FUN=mean, na.rm=TRUE)
aggregate(data=X, cbind(links, mentions, filesize, approxTweets, norm, fact) ~ dow, FUN=mean, na.rm=TRUE)

# Est. tweets
p <-ggplot(aes(dates, approxTweets), data=X) +
        geom_line(color="blue") +
	coord_flip()+
	geom_smooth() + 
	xlab("Day") + ylab("Tweets/Hour (est. on file size)") +
	scale_x_datetime(major = "2 days", format = "%d") 
ggsave(file="./tweets.png", width=6, height=12, dpi=100)

m <- acf(X$norm)
lag.plot(X$norm, 24) 

p <-ggplot( data=X) +
        geom_boxplot(aes(dow, approxTweets), color="blue") +
	coord_flip()+
        xlab("Day of Week") + ylab("Tweets per Hour") +
ggsave(file="./dow.png", width=6, height=6, dpi=100)

# hours
p <-ggplot( data=X) +
        geom_boxplot(aes(hour, approxTweets), color="blue") +
	coord_flip()+
        xlab("Hour of Day") + ylab("Tweets per Hour") +
ggsave(file="./hour.png", width=6, height=9, dpi=100)