Contents Index <<Browse<< >>Browse>>
Proc spectra is used to identify periodicities in time series data. The data sets analyzed here are the airline passenger miles data and the milk production data used in Project 11. The raw data is found in the files airline.dat and milk.dat.
Suggestion: If you plan to use high quality graphs in your work, read about proc gplot and do Project 7 before doing this assignment.
1. Use proc spectra to find the periodogram of the milk production data. Since the data is monthly, it should be no surprise to find the data is periodic with period 12. How is this seen from your analysis? What does the test for white noise show for the original data? The periodicity suggests differencing the original series at lag 12 should produce a stationary time series. Create a new data set containing the differenced series. You may want to use the dif operator. Are there any remaining periodicities in the difference series? Is the new series white noise? Choose appropriate weights and compute the spectral density estimate of the differenced series. Turn in your a nice copy of your program and submit nice print outs from SAS to back up your answers to the questions above. There is no need to turn in a printed copy of the data.
2. Repeat the analysis of problem 1 using the airline data. Recall that a logarithmic transform of the raw data is suggested before analysis. Are there periodicities? How can they be removed? Submit nice print outs of your program and analysis from SAS to back up your answers. There's no need to print out the data.
3. If you can use proc gplot, make some nice graphs of the periodogram and spectral density estimates to supplement your analysis in the preceding problems and turn them in also.
Contents Index <<Browse<< >>Browse>>
Copyright © 1997 by Jerry Alan Veeh. All rights reserved.