Time Series Assignment
Rubric
Refer to the spreadsheet ‘ACF702 Assignment Timeseries Data Referral’ and choosing either dataset A or B.
- Define, estimate and show the ACF and PACF for your time series. Interpret the results. (10 marks)
- Use Augmented Dickey-Fuller tests to determine whether your chosen series may contain a unit root process. Transform and re-test as appropriate to
determine the order of integration.Show your test output. (20 marks)
- Select the most appropriate ARIMA time series model for your data on the basis of ACF and PACF plots, statistical tests on estimated coefficients / parameters, information criteria and appropriate experimentation. Show the output of these tests. Save and display the estimated parameter values of the two equations you feel best fit the data and indicate which is you preferred NOTE: You are advised to estimate these models in Gretl using the
‘Conditional Maximum Likelihood’ option within the ARIMA time series options
(40 marks)
- Withhold five per cent of the most recent data and re-estimate the two equation functional forms you selected in (3). Using the re-estimated equations calculate forecasts (for the withheld data) for each model using
Excel. Judge which forecasts are ‘best’ using a criterion such as RMSE.
(30 marks)
Additional information
All key terms, concepts and methods should be defined and explained throughout. No marks will be given for merely presenting results without explanation. In particular, be extremely clear in how you have calculated the forecasts (e.g. provide workings for at least the first two forecasts). You should reflect and comment on your analysis.
Submitted work must not be bound with a spine but should be stapled once in the top left hand corner. You must state which dataset you have chosen on the front cover.
Word Limit: 1500 words