Project #1 Guidelines
General Information
• Deadline: October 28, 11:59 pm
• Work Independently: keep your do/Rmd files and your dataset private from other students.
Copying and modifying another student’s code is plagiarism.
• Length of the project: up to 5 pages of text, not including figures and tables
• Your paper should be typed and spell-checked
• Use ARIMA methods to forecast and analyze a time series of your choice o Minimum number of observations: 50 o Time series should be different from the ones used in lectures and problem sets. o The last observation should be recent and must be the most up-to-date observation available from the source.
• Structure your project like a paper:
o Start with a short introduction describing the data source and series you are analyzing, your motivation for choosing this series, and related literature (if relevant)
o Describe the analysis you performed and the results
o All the main results should be reported in tables (you cannot copy the STATA/R output
– look up real papers to see how tables should look) o Write a conclusion summarizing your main findings
o The quality of writing and formatting will matter for the grade
Elements to include in your analysis
• Plot the data; describe the patterns you see; decide whether transformation is needed.
• If the data are seasonal, then it would be best to remove this component
• Analyze it for stationarity
• Fit several appropriate ARIMA models using the ACF, PACF, and information criteria to select an ARIMA model.
• Present the parameter estimates for the best model. Do not give the estimates for the other models you tried. A table with information criteria for all fitted models is sufficient.
• Investigate structural breaks
• Construct a forecast (be specific about the date for which you make the forecast, and the method used for forecasting) and plot the data, together with the forecast and the 95% forecast intervals. Comment briefly on whether the forecasts seem reasonable.