04.01.2013 Views

Springer - Read

Springer - Read

Springer - Read

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

168 Chapter 5 Modeling and Forecasting with ARMA Processes<br />

are given by<br />

P57X57+h −4.035 +<br />

1<br />

jh<br />

θ57+h−1,j<br />

<br />

X57+h−j − ˆX57+h−j<br />

⎧<br />

<br />

⎨ −4.035 + θ57,1 X57 − ˆX57 ,<br />

<br />

⎩<br />

−4.035,<br />

with mean squared error<br />

if h 1,<br />

if h>1,<br />

<br />

2040.75r57, if h 1,<br />

E(X57+h − P57X57+h) 2 <br />

2040.75(1 + (−.818) 2 ), if h>1,<br />

where θ57,1 and r57 are computed recursively from (3.3.9) with θ −.818.<br />

These calculations are performed with ITSM by fitting the maximum likelihood<br />

model (5.4.1), selecting Forecasting>ARMA, and specifying the number of forecasts<br />

required. The 1-step, 2-step, ...,and 7-step forecasts of Xt are shown in Table 5.1.<br />

Notice that the predictor of Xt for t ≥ 59 is equal to the sample mean, since under<br />

the MA(1) model {Xt,t ≥ 59} is uncorrelated with {Xt,t ≤ 57}.<br />

Assuming that the innovations {Zt} are normally distributed, an approximate 95%<br />

prediction interval for X64 is given by<br />

−4.0351 ± 1.96 × 58.3602 (−118.42, 110.35).<br />

The mean squared errors of prediction, as computed in Section 3.3 and the example<br />

above, are based on the assumption that the fitted model is in fact the true model<br />

for the data. As a result, they do not reflect the variability in the estimation of the<br />

model parameters. To illustrate this point, suppose the data X1,...,Xn are generated<br />

from the causal AR(1) model<br />

Xt φXt−1 + Zt, {Zt} ∼iid 0,σ 2 .<br />

Table 5.1 Forecasts of the next 7 observations<br />

of the overshort data of Example<br />

3.2.8 using model (5.4.1).<br />

# XHAT SQRT(MSE) XHAT + MEAN<br />

58 1.0097 45.1753 −3.0254<br />

59 0.0000 58.3602 −4.0351<br />

60 0.0000 58.3602 −4.0351<br />

61 0.0000 58.3602 −4.0351<br />

62 0.0000 58.3602 −4.0351<br />

63 0.0000 58.3602 −4.0351<br />

64 0.0000 58.3602 −4.0351

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!