arimaPredict#

Purpose#

Estimates forecasts using estimation results obtained from arimaFit().

Format#

f = arimaPredict(b, y, e, h)#

Parameters:

b (Kx1 vector) – estimated coefficients
y (Nx1 vector) – data.
e (Nx1 vector) – residuals reported by arimamt program.
h (scalar) – the number of step-ahead forecasts to computer.

Returns:

f (hx3 matrix) –

\([.,1]\)	Lower forecast confidence bounds.
\([.,2]\)	Forecasts.
\([.,3]\)	Upper forecast confidence bounds.

Example#

new;
cls;
library tsmt;

// Simulate data
seed = 423458;
y = simarmamt(.3, 1, 0, 2, 0, 250, 1, .5, seed);

// Integrated series
z = cumsumc(y);

// Declare arima out structures
struct arimamtOut amo;

// Set AR order
p = 1;

// Set order of differencing
d = 1;

// Estimate model
amo = arimaFit(z, p, d);

// Forecast model
f = arimaPredict(amo, z, 25);

The first five forecasts printed to the screen are:

Forecasts for ARIMA(1,1,0) Model.   95% Confidence Interval Computed.

Period      LCL        Forecasts        UCL     Forecast Std. Err.
  480.694133    481.751067    482.808002      0.539262
  481.268540    483.021519    484.774499      0.894394
  482.000099    484.313042    486.625985      1.180095
  482.830115    485.611374    488.392633      1.419036
  483.724926    486.911907    490.098888      1.626041

Remarks#

Data must be transformed before being sent to arimaPredict().
The arimaPredict() procedure does not compute forecasts for models with fixed regressors.

Library#

tsmt

Source#

forecastmt.src