THE UNIVERSITY OF HONG KONG 20/21
DEPARTMENT OF STATISTICS AND ACTUARIAL SCIENCE
STAT4601 Time Series Analysis
Project
This is an individual project, and each of you is expected to submitted a report before
the end of this semester (to be determined).
1 Guidelines for the project
Find a time series with sample size n > 100, and delete the last 5 values for the sake of
comparison at the end of this project. Answer the following questions one by one, and the
SAS, R or other suitable software is needed to derive the answers.
1. Introduce briefly the background of this time series.
2. Draw the time plot, and check whether or not it is stationary and whether or not it
needs transformations.
3. Transfer this time series into a stationary sequence if possible. Specify the reasons to
your action.
4. Specify the model(s) based respectively on sample ACFs, and on the sample PACFs.
5. For each specified model,
(a) estimate the parameters by the maximum likelihood estimation method.
(b) use Ljung-Box test to check the fitted model is adequate or not. Please try several
different K here.
(c) use overparameterized method to check the adequacy of the fitted model.
(d) suggestion some alternative models if the fitted model is not adequate.
(e) check these alternative models.
6. Use AIC to select one model if there are more than one adequate models.
7. Forecast the 5 future values, and compare them with the true values.
8. Any new insight into this time series based on the above process?
Note: the above listed items are ONLY for your reference, and you do not need
to exactly follow them.
1