A) Plot the data in 1982 dollars and find the least-squares trend line. What fraction of the variability in revenue is accounted for by the trend alone?
B) Find the quarterly seasonal indices for the revenues in 1982 dollars, and use them to deseasonalize those revenues.
C) Find the least-squares trend line for the deseasonalized data.
D) 1) Use the deseasonalized trend line to predict the deseasonalized revenues for all 40 quarters.
2) Reseasonalize those forecasts by multiplying by the appropriate seasonal indices and dividing by 100.
3) For each of the 40 quarters, subtract the actual revenue from the reseasonalized forecast to find the error in the forecast.
4) Square these errors and sum them up. Call the result SSE*
5) Let SST be the total sum of the squares for the trend line in part (A). The fraction of the variation in the actual revenues explained by the trend and the seasonality is 1-SSE*/SST. How much more of the variability in the quarterly revenues is explained by taking the seasonality into account?
See attached file for full problem data.
This posting contains solution to following timeseries forecasting problem with seasonality.