S-Plus commands used in time series analysis
==============================================

** "%" denotes explanation
I used monthly U.S. unemployment rate series as an illustration.


x=read.table(file="m-unrate.txt")   % Load data into S-Plus workspace

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

plot(rate,type='l')  % Plot the time series.

zt = diff(rate)    % Take the first difference

t.test(zt)       % Perform t-test to test E(zt) = 0.

acf(zt,lag.max=15) % Obtain the first 15 lages of ACF

acf(zt,type="partial",lag.max=15) % Obtain the first 15 lags of PACF.

m1=arima.mel(zt, model=list(order=c(2,0,1)))   % fit an ARIMA(2,0,1) model

summary(m1)   % Obtain the fitted results

m1chk = arima.diag(m1)  % Perform model checking and obtain std. residuals

plot(m1)  % obtain plots of model checking (similar to arima.diag)

names(m1chk)  % Produce what are contained in the object "m1chk".

predict.arima(m1,n.predict=5) % produce 1- to 5-step ahead forecasts.

Remarks:
========
(1) If you load the module "finmetrics", then you can use the 
command "autocorTest" to perform Q-test of ACF.
(2) In S-Plus, arima.mle assumes zero mean of the time series. 
    If the mean is not zero, you can do two things:
(a) Create a regressor of "1" to add to the model.

For example, the differenced unempolyment rate series has 691 observations.
z1=rep(1,691)

m1=arima.mle(zt,model=list(order=c(2,0,1)),xreg=z1)

(b) Remove the sample mean before fitting the model.

ztm = zt-mean(zt).

END of Remarks
==============