You will be supplied by e-mail with a set of data. It is a quarterly series of sales from the first quarter 1993 to the second quarter 2008. It is not seasonally adjusted.
1. Examine the data, looking for seasonal effects, trends and cycles. (10 marks)
2. Choose some of the data to model and some to use as a check on the forecasts you will be producing. Explain your choice. (5 marks)
3. Try a Classical Decomposition method on the chosen data for modelling, comment on the analysis and check the forecasts against the rest of the data. (15 marks)
4. Try a regression with Dummy Variables method with a linear trend component on the same data. Comment on the analysis and check the forecasts against the rest of the data (15 marks)
5. Try a Lagged Variables approach on the same data. Comment on the analysis and check the forecasts against the rest of the data (15 marks)
6. Remove the seasonal effect first using the classical decomposition method and then try the Box-Jenkins ARIMA approach. Comment on the analysis and compare your forecasts with the holdback data. (25 marks)
7. Discuss all the differences between the results from the different methods and include where possible comments on the different form of the equations for the forecasts from the different methods you have used. Choose your final model and briefly explain your choice (15 marks)
/* Style Definitions */
mso-padding-alt:0cm 5.4pt 0cm 5.4pt;
font-family:”Times New Roman”,serif;}
Your answer should be in the form of a structured report (i.e. including appropriate sections and headings). In total your written material, including SPSS output should not exceed 18 pages (not including the title page). Use a font size of 11 or 12. It is recommended that you put more than one set of annotated SPSS results on a page to keep within the limit.