### This file contains some examples from Ch. 1 Splus

# to read jj.dat 

x <- scan("E:\\Stat207\\Class\\Examples\\Ch1\\jj.dat")

# you can make x a regular time series object

x <- rts(x)

# plot the data

ts.plot(x, xlab="Quarter", ylab="J&J Earning Per Share")

# log and plot

x.log <- log(x)
ts.plot(x.log, xlab="Quarter", ylab="Log of J&J Earning Per Share")

# difference the log data and plot

dx <- diff(x.log)
ts.plot(dx, xlab="Quarter", ylab="J&J Earning Per Share Growth Rate")

# all 3 plots - one on top of the other
par(mfrow=c(3,1)) 
ts.plot(x, xlab="Quarter", ylab="J&J Earning Per Share")
ts.plot(x.log, xlab="Quarter", ylab="Log of J&J Earning Per Share")
ts.plot(dx, xlab="Quarter", ylab="J&J Earning Per Share Growth Rate")

# Detrend & Regression        
z <- 1:length(x.log)
fit <- lm(x.log~z)

par(mfrow=c(1,1))
min(x.log)
max(x.log)
ts.plot(x.log, xlab="Quarter", ylab="Log of J&J Earning Per Share",ylim=c(-1,3) )
par(new=T)
x.hat <- fit$fitted
tsplot(x.hat, xlab="", ylab="", ylim=c(-1,3))  # both lines on plot

#detrended series
dt <- x.log-x.hat     
ts.plot(dt)

# compare difference with detrended data
par(mfrow=c(2,1))
acf(dx)           # remember dx <- diff(x.log)
acf(dt)

# matrix of plots of dx(t) vs dx(t-h) for h=1,2,...,9
lag.plot(dx, lags=9, layout=c(3,3)) 

# dummy variable regression
u <- matrix(contr.sum(4, contrasts=F),4,84)  # matrix of dummy variables
u <- t(u)               # transpose
fit <- lm(x.log~z + u[,1:3])     # fit on 3 cols of dummy
summary(fit)   # get details of the regression
noise <- fit$residuals  # get residuals  
acf(noise)
tsplot(noise)