HeathCo Industries, a producer of a line of skiwear, has been the subject of exercises in several earlier chapters of the text. The data for its sales and two potential causal variables, income (INCOME) and the northern-region unemployment rate (NRUR), are repeated in the following table:

a. Develop a multiple-regression model of SALES as a function of both INCOME and NRUR:

SALES = a + b1 (INCOME) + b2(NRUR)

Use this model to forecast sales for 2008Q1-2008Q4 (call your regression forecast series SFR), given that INCOME and NRUR for 2004 have been forecast to be:

b. Calculate the RMSE for your regression model for both the historical period (1998Q1-2007Q4) and the forecast horizon (2008Q1-2008Q4).

PeriodRMSE

Historical€¦€¦€¦€¦€¦€¦..________

Forecast€¦..€¦€¦€¦€¦€¦..________

c. Now prepare a forecast through the historical period and the forecast horizon (2008Q1-2008Q4) using Winters' exponential smoothing. Call this forecast series SFW, and fill in the RMSEs for SFW:

PeriodRMSE

Historical€¦€¦€¦€¦€¦€¦..________

Forecast€¦€¦€¦€¦€¦€¦€¦________

d. Solely on the basis of the historical data, which model appears to be the best? Why?

e. Now prepare a combined forecast f SCF) using the regression technique described in this chapter. In the standard regression:

SALES = a + b1 (SFR) + b2(SFW)

Is the intercept essentially zero? Why? If it is, do the following regression as a basis for developing SCF:

SALES = b1 (SFR) + b2(SFW)

Given the historical RMSEs found in parts (b) and (c), do the values for b and b2 seem plausible? Explain.

f. Calculate the RMSEs for SCF:

PeriodRMSE

Historical€¦€¦€¦€¦€¦€¦..________

Forecast€¦€¦€¦€¦€¦€¦€¦________

Did combining models reduce the RMSE in the historical period? What about the actual forecast?

Transcribed Image Text:

ur 52 125 99986 6754 8653836 0522 5 8272 271 990 677 1582 3 830907094 2991 a-3724996223541 48r S-550228 321 2222222233323 7