Distribution and Inventory Management
| Period t | Year | Quarter | Demand |
| 1 | 1 | I | 46 |
| 2 | II | 43 | |
| 3 | III | 56 | |
| 4 | IV | 62 | |
| 5 | 2 | I | 65 |
| 6 | II | 69 | |
| 7 | III | 83 | |
| 8 | IV | 79 | |
| 9 | 3 | I | 83 |
| 10 | II | 75 | |
| 11 | III | 92 | |
| 12 | IV | 92 |
a. four-period moving average (start with forecast for period 5)
b. exponential smoothing with alpha=0.1
c. exponential smoothing with alpha=0.6
d. Holt’s model with alpha=0.2 and beta=0.4, and
e. a regression line through the data (as we did in static forecasting, use the functions =INTERCEPT() and =SLOPE()).
3. For a-e as given in problem 2, calculate for MAD, MSE, bias, and TS, for periods 5-12, and comment on the following:
-
The difference between moving average and exponential smoothing.
-
The difference between level-only models (a, b, c) and level+trend models (d and e).
-
The difference between Holt’s model and the basic regression line. Why do you think most ‘real-world’ forecasters prefer a model like Holt’s to a regression? Think about what is desired in forecasting.
SAMPLE ASSIGNMENT
