Contents Index <<Browse<< >>Browse>>
Some time domain methods (Box-Jenkins methods) for the analysis of time series using proc arima are used to analyze 2 sets of data.
The first data set, milk.dat, contains milk production data by month for the years 1962-1975 (Cryer). The second data set, airline.dat, contains the yearly millions of air passenger miles flown each month within the United States for the years 1960-1977 (Cryer).
Suggestion: If you plan to use high quality graphs in your work, read about proc gplot and do Project 7 before starting this assignment.
1. Make a SAS data set containing the milk production data for the years 1962-1974, leaving the 1975 data out of the analysis. You don't need to turn in a print out of the data set. Use proc arima to fit an ARIMA model to the time series and use this model to predict milk production for 1975. Print out a copy of the final analysis and forecasts for 1975 using the model you select. Based on the printout, how adequate is your model for the data? Turn in a copy of your program too.
2. Conduct an analysis of the airline data similar to that of problem 1, leaving the 1977 data out of the analysis. The raw airline data shows increasing variablility with trend, so the data should be transformed prior to analysis in order to obtain a stationary time series. Add a variable logmiles to the data set containing the raw airline data, and conduct the analysis using proc arima on this new variable. Remember to exponentiate the forecasts from this analysis to obtain forecasts of air passenger miles.
3. If you are using proc gplot to produce high quality graphs, make graphs of the original milk data and its sample autocorrelation function. Do the same for the raw airline data, the log of the airline data, and the autocorrelation function of the log of the airline data.
Contents Index <<Browse<< >>Browse>>
Copyright © 1997 by Jerry Alan Veeh. All rights reserved.