ANL317: Discuss the Importance of Stationarity for Forecasting and How to Convert a non-Stationary Series to Become Stationary: Business Forecasting Assignment, SUSS, Singapore

Question 1

Stationarity is one of the main time-series properties.

• Discuss the importance of stationarity for forecasting and how to convert a non-stationary series to become stationary.
• Give an example of a non-stationary time series. Explain the difficulty and possible adjustment measures in doing forecasting on your example series.

Question 2

Below are the quarterly gross domestic product (GDP) data of country X from 1980 to 1989.

Create a SAS Studio dataset/SAS Forecast Studio project for the analytics tasks below.

a. Determine the trend of the series by applying linear regression for curve fitting.

b. Evaluate whether there are any seasonal effects in this time series by creating a new quarterly variable as the independent variable of linear regression.

c. Combine the trend and seasonal variables that you have used in (a) and (b) to construct a joint linear regression model for the GDP. Comment on whether the findings regarding the trend and seasonal effects in (a) and (b) are still relevant in this joint effect model.

d. Compute the simple moving averages for these data with the most appropriate order by comparing the mean absolute percentage errors (MAPE) of some possible moving average orders.

e. Propose and construct the most suitable exponential smoothing model for these data based on the findings in (a), (b), and (c). Report and conclude on the coefficients that you can find in the output of SAS Forecast Studio.

f. Calculate the GDP forecast for the four quarters in 1990 using the models estimated in (c), (d) and (e). Comment on the forecasting accuracy of these models if the observed data are