Complete the given questions.

Data sets for the assignment is given.

ASSIGNMENT: Forecasting Trends: Exponential Smoothing

INSTRUCTIONS: Use MINITAB to complete the following questions.

1. Use the data set BND.MTW. These data show annual U.S. bond yield from 1900 to 1970. The source for these data is Hyn¬d¬man, R.J. TimeSeries Data Library.

(a) Prepare a time series plot for these data. What is (are) the dominant time series component(s)?

(b) Fit a moving average of order 1 (naïve), order 3 and order 6 to these data. Based on the fitted accuracy measures, which works best?

(c) Fit a single exponential smoothing model (with an optimal smoothing parameter value) to these data. How does it compare with the moving average models fit in part (a)?

2. Use the data set INTERNETUSAGE.MTW. The file contains yearly data on the percentage of individuals using the Internet in Germany from 1990 to 2011 (Source: United Nations).

(a) Use Holt's Method (double exponential smoothing) to fit a model to these data (use the default - optimal smoothing constants).

(b) Fit a linear trend model to these data (this is part of decomposition). To do this go to Time Series > Trend Analysis and check Linear.

(c) Fit the S-curve trend model to these data (this is part of decomposition).  To do this go to Time Series > Trend Analysis and check S-Curve(Pearl-Reed logistic).

(d) Explain why these three approaches are appropriate choices for these data.

(e) The United Nations determined that the percentage of individuals using the Internet in 2012 was 84%. Use each of the three models fit above to forecast the percentage for 2012. Which method gives the best forecast?
 
3. Use the data set FedReceipt.MTW provided with this assignment. These data represent federal receipts from 1978 through 2011, in billions of dollars, from individual and corporate income tax, social insurance, excise tax, estate and gift tax, customs duties and federal reserve deposits (Source: bit.ly/7dGCmz)

(a) Fit a linear, quadratic and exponential trend model to these data. Based on the fitted accuracy measures, which works best?

(b) The value for receipts in 2012 is 2450.2 billion dollars. Use the three different trend models fit in part (a) to forecast the value for 2012. Which one forecasts the best?

Attachment:- Minitab.rar