The ARIMA Procedure

Example 8.3 Model for Series J Data from Box and Jenkins

This example uses the Series J data from Box and Jenkins (1976). First, the input series X is modeled with a univariate ARMA model. Next, the dependent series Y is cross-correlated with the input series. Since a model has been fit to X, both Y and X are prewhitened by this model before the sample cross-correlations are computed. Next, a transfer function model is fit with no structure on the noise term. The residuals from this model are analyzed; then, the full model, transfer function and noise, is fit to the data.

The following statements read Input Gas Rate and Output CO $_{2}$ from a gas furnace. (Data values are not shown. The full example including data is in the SAS/ETS sample library.)

title1 'Gas Furnace Data';
title2 '(Box and Jenkins, Series J)';
data seriesj;
   input x y @@;
   label x = 'Input Gas Rate'
         y = 'Output CO2';
datalines;
-0.109  53.8  0.000  53.6  0.178  53.5  0.339  53.5

   ... more lines ...

The following statements produce Output 8.3.1 through Output 8.3.11:

proc arima data=seriesj;

   /*--- Look at the input process ----------------------------*/
   identify var=x;
   run;

   /*--- Fit a model for the input ----------------------------*/
   estimate p=3 plot;
   run;

   /*--- Cross-correlation of prewhitened series ---------------*/
   identify var=y crosscorr=(x) nlag=12;
   run;

   /*--- Fit a simple transfer function - look at residuals ---*/
   estimate input=( 3 $ (1,2)/(1) x );
   run;

   /*--- Final Model - look at residuals ----------------------*/
   estimate p=2 input=( 3 $ (1,2)/(1) x );
   run;

quit;

The results of the first IDENTIFY statement for the input series X are shown in Output 8.3.1. The correlation analysis suggests an AR(3) model.

Output 8.3.1: IDENTIFY Statement Results for X

Gas Furnace Data

(Box and Jenkins, Series J)

The ARIMA Procedure

Name of Variable = x
Mean of Working Series	-0.05683
Standard Deviation	1.070952
Number of Observations	296

Output 8.3.2: IDENTIFY Statement Results for X: Trend and Correlation

The ESTIMATE statement results for the AR(3) model for the input series X are shown in Output 8.3.3.

Output 8.3.3: Estimates of the AR(3) Model for X

Conditional Least Squares Estimation
Parameter	Estimate	Standard Error	t Value	Approx Pr > \|t\|	Lag
MU	-0.12280	0.10902	-1.13	0.2609	0
AR1,1	1.97607	0.05499	35.94	<.0001	1
AR1,2	-1.37499	0.09967	-13.80	<.0001	2
AR1,3	0.34336	0.05502	6.24	<.0001	3

Constant Estimate	-0.00682
Variance Estimate	0.035797
Std Error Estimate	0.1892
AIC	-141.667
SBC	-126.906
Number of Residuals	296

Model for variable x
Estimated Mean	-0.1228

Autoregressive Factors
Factor 1:	1 - 1.97607 B(1) + 1.37499 B(2) - 0.34336 B**(3)

The IDENTIFY statement results for the dependent series Y cross-correlated with the input series X are shown in Output 8.3.4, Output 8.3.5, Output 8.3.6, and Output 8.3.7. Since a model has been fit to X, both Y and X are prewhitened by this model before the sample cross-correlations are computed.

Output 8.3.4: Summary Table: Y Cross-Correlated with X

Correlation of y and x
Number of Observations	296
Variance of transformed series y	0.131438
Variance of transformed series x	0.035357

Both series have been prewhitened.

Output 8.3.5: Prewhitening Filter

Autoregressive Factors
Factor 1:	1 - 1.97607 B(1) + 1.37499 B(2) - 0.34336 B**(3)

Output 8.3.6: IDENTIFY Statement Results for Y: Trend and Correlation

Output 8.3.7: IDENTIFY Statement for Y Cross-Correlated with X

The ESTIMATE statement results for the transfer function model with no structure on the noise term are shown in Output 8.3.8, Output 8.3.9, and Output 8.3.10.

Output 8.3.8: Estimation Output of the First Transfer Function Model

Conditional Least Squares Estimation
Parameter	Estimate	Standard Error	t Value	Approx Pr > \|t\|	Lag	Variable	Shift
MU	53.32256	0.04926	1082.51	<.0001	0	y	0
NUM1	-0.56467	0.22405	-2.52	0.0123	0	x	3
NUM1,1	0.42623	0.46472	0.92	0.3598	1	x	3
NUM1,2	0.29914	0.35506	0.84	0.4002	2	x	3
DEN1,1	0.60073	0.04101	14.65	<.0001	1	x	3

Constant Estimate	53.32256
Variance Estimate	0.702625
Std Error Estimate	0.838227
AIC	728.0754
SBC	746.442
Number of Residuals	291

Output 8.3.9: Model Summary: First Transfer Function Model

Model for variable y
Estimated Intercept	53.32256

Input Number 1
Input Variable	x
Shift	3

Numerator Factors
Factor 1:	-0.5647 - 0.42623 B(1) - 0.29914 B(2)

Denominator Factors
Factor 1:	1 - 0.60073 B**(1)

Output 8.3.10: Residual Analysis: First Transfer Function Model

The residual correlation analysis suggests an AR(2) model for the noise part of the model. The ESTIMATE statement results for the final transfer function model with AR(2) noise are shown in Output 8.3.11.

Output 8.3.11: Estimation Output of the Final Model

Conditional Least Squares Estimation
Parameter	Estimate	Standard Error	t Value	Approx Pr > \|t\|	Lag	Variable	Shift
MU	53.26304	0.11929	446.48	<.0001	0	y	0
AR1,1	1.53291	0.04754	32.25	<.0001	1	y	0
AR1,2	-0.63297	0.05006	-12.64	<.0001	2	y	0
NUM1	-0.53522	0.07482	-7.15	<.0001	0	x	3
NUM1,1	0.37603	0.10287	3.66	0.0003	1	x	3
NUM1,2	0.51895	0.10783	4.81	<.0001	2	x	3
DEN1,1	0.54841	0.03822	14.35	<.0001	1	x	3

Constant Estimate	5.329425
Variance Estimate	0.058828
Std Error Estimate	0.242544
AIC	8.292809
SBC	34.00607
Number of Residuals	291

Output 8.3.12: Residual Analysis of the Final Model

Output 8.3.13: Model Summary of the Final Model

Model for variable y
Estimated Intercept	53.26304

Autoregressive Factors
Factor 1:	1 - 1.53291 B(1) + 0.63297 B(2)

Input Number 1
Input Variable	x
Shift	3

Numerator Factors
Factor 1:	-0.5352 - 0.37603 B(1) - 0.51895 B(2)

Denominator Factors
Factor 1:	1 - 0.54841 B**(1)