The Sales2 variable in the file from the previous problem was created from the Sales1 variable by multiplying by monthly seasonal factors. Basically, the summer months are high and the winter months are low. This might represent the sales of a product that has a linear trend and seasonality.
a. Repeat parts a, b, and c from the previous problem to see how well these forecasting methods can deal with trend and seasonality.
b. Now use Winters’ method, with various values of the three smoothing constants, to forecast the series. Can you do much better? Which smoothing constants work well?
c. Use the ratio-to-moving-average method, where you first de-seasonalize the series and then forecast (by any appropriate method) the de-seasonalized series. Does this perform as well as, or better than, Winters’ method?
d. What can you conclude from your findings in parts a, b, and c about forecasting this type of series?

  • CreatedApril 01, 2015
  • Files Included
Post your question