The data presented in the datafile Quarterly Appliance Sales are suitable for developing models for predicting the four kinds of appliance sales. Three kinds of models can be developed. These are:
1. A model with appropriate quarterly indicators, a time variable, coincident and lagged values of DUR or RES, and appropriate lagged values of the dependent variable.
2. A model using only predictor variables which lead the dependent appliance series (delete from #1 coincident values of DUR or RES).
3. A model using only appropriate seasonal indicators and lagged values of the dependent appliance series (delete from #2 lagges values of DUR or RES).
For each model type considered, create a suitable model and test the resulting residuals for independence with time. If the residuals fail the test of independence with time, you may wish to add lagged values of residuals to your model. Compare the standard errors of estimate from your various models.
For comparing forecasts with actual values, the values of the apppliance sales series for 1986 are
Quarter DISH DISP FRIG WASH
1 898 1063 1274 1297
2 955 1028 1822 1399
3 1008 1073 1932 1575
4 1054 1089 1494 1495