Some R commands for time series analysis
========================================

package: tseries

Use the command

setwd("C:/....") to set your working directory.

x=read.table("m-unrate.txt") % Load data into R

rate=x[,4]  % Get the unemployment rate series

ts.plot(rate) % Plot the time series.

** There are two commands to fit an ARMA models:
** ``arma'' and ``arima''. Each has its own features. 
** For forecasting purpose, "arima" is better.

** ARMA command **

zt=diff(rate) % Take the first difference

acf(zt,lag.max=15) % Compute the first 15 ACFS

pacf(zt, lag.max=15) % Compute first 15 lags of PACFs

m1=arma(zt, order=c(2,1)) % Fit an ARMA(2,1) model.

summary(m1)  % obtain fitted results

names(m1) % provide a list of objects stored in "m1".

ts.plot(m1$residuals) % obtain time plot of residuals

plot(m1) % Obtain several plots. Click onthe plot window to see next plots.

acf(m1$residuals,lag.max=15) % comput residuals ACFs

Remark:
======
The command "arma" fits a constant term as a default option. You can 
include include.intercept=F to drop the constant term if needed.

End of Remark.
=============

** ARIMA command **

You can use the command "arima" to fit the model as well.
The order will be in the form 

order=c(p,d,q). 

If you use this command, then you can perform model checking via 
the command "tsdiag" and compute forecasts via the command "predict".

For example,
  m3=arima(zt,order=c(2,0,1)) % fit an ARIMA(2,0,1) model to zt.
  tsdiag(m3)    % model checking
  predict(m3,n.ahead=5) % produce 1- to 5-step ahead forecasts.

Or more directly,
 m4=arima(rate,order=c(2,1,1))
 tsdiag(m4)
 predict(m4,n.ahead=5)