Principals of Forecasting and Applications
The unit learning objectives of this assignment are:
• Motivate the need for obtaining reliable forecasts in business and economics
• Understand and apply appropriate statistical methods for business and economic forecasting
• Develop computer skills for forecasting from business and economic time series data • Provide practical insights from your forecasts
Add
INSTRUCTIONS
4. Your report should be no more than 15 pages (excluding Graphs, Reference List, and Appendix). Any part of the report beyond the 15 page limit will be struck out and not marked.
(a) Use default format, paragraph, and margin settings.
(b) Font size: 12
(c) At least 1.2 line spacing between lines.
(d) Graphs should be appropriately sized and easy to read. They should not be made small to conserve space.
5. With regards to graphs and estimation outputs:
(a) All graphs should be in-line with the text for ease of reading, and not placed in an Appendix at the end of the report.
(b) Any raw R output can be labelled and placed in an Appendix at the end of the report. Otherwise, if reporting any estimation output within the report, the output should be professionally presented in a table format.
7. All submissions will be via Moodle.
(a) If you choose to type your assignment in Microsoft Word, you will need to save it as a PDF file and submit (i) the PDF document, (ii) the Excel dataset (.csv format), and (iii) the R-script file consisting of the codes used to perform your analysis.
(b) If you choose to type your assignment using Rmarkdown, you will need to submit (i) the PDF form of the assignment, (ii) the Excel dataset (.csv format), and (iii) the RMD file.
8. All submissions should be submitted with an Assignment Cover Sheet attached.
Add
ASSIGNMENT TASK
Assignment Aim
Use the ARIMA framework to quantify the forecasted loss in US revenue passenger miles since the COVID-19 pandemic.
Instructions
For this assignment, write a research report based on your analysis. Your report should include tables, graphs, and associated narrative. Keep the report concise and clear. Thought- fulness, clarity of your discussion and the communication of your results are important. The Appendix does not constitute part of the 15-page limit.
Add
Method
The Box Jenkins methodology is a popular tool used in forecasting. Using the revenue passenger miles data from your Group Assignment 1, your task is to identify, estimate and select the best forecasting method from the ARIMA class, using the software R.
Your report should clearly highlight and explain the following phases of the Box Jenkins methodology:
Data Preliminaries
• For this Assignment, use the full dataset that you have obtained from “Stage 4: PreProcess Data” of Group Assignment 1.
• As with Stage 5 of Group Assignment 1, partition your full dataset into 2 parts labelled as “Pre-Covid” and “Covid”.
– The “Pre-Covid” part comprises of data prior to the start of the pandemic in the United States.
– Ensure that the “Covid” part comprises of data that coincides with the time frame when the United States was affected by the pandemic.
• Write a brief introduction to the report before commencing the tasks below. It shouldinclude the contextual background of the assignment.
• Using the “Pre-Covid” dataset, partition your data into a training and test set, specifying the time duration over which the respective sets span. (The training and testAdd sets should be the same as your Group Assignment 1, unless they were commented as unsuitable in the feedback).
• Plot the training set data and determine if any transformation is necessary.
• If deemed necessary, transform the data using appropriate tools and techniques prior to analysis, taking heed to provide the relevant details.
• Using appropriate graphic plots for the (transformed) training set data, identify at least TWO possible ARIMA models for subsequent comparison purposes. Also, determine and explain if your specified ARIMA models should have a constant.
• Be sure to articulate the process and the reasoning behind the identification of the models.
• Estimate your identified ARIMA models from Phase 1 using the training set data.
• Estimate another ARIMA model using the auto.arima() function. Set the stepwise argument as FALSE.
• Report the parameter estimates of the ARIMA models, along with other relevant statistics, in a table.
• Rank the ARIMA models according to an appropriate information criterion. We will proceed to compare your top 2 models.
• Perform diagnostic checks on the residuals for these 2 models. Determine if both models are adequate.
• Does any of your 2 models need to be re-identified? In other words, if the model residuals are not white noise, try a modified model by modifying the values of theAR and/or MA component. If your re-identified model still does not return a white noise residual, comment that the model does not return a white noise residual process and proceed on with the rest of the assignment. Note that you MUST articulate anddemonstrate this process, and not reveal your final 2 models without first showing how you obtained them.
• Determine which of these 2 models is a better forecasting model and write out thisAdd model using backshift notation.
• Re-estimate your selected model from Phase 2 using the Pre-Covid data. Report the estimated model using the backshift notation.
• Produce forecasts for the “Covid” time period.
• Produce a plot of the full dataset (i.e., “Pre-Covid” and “Covid”) with the forecasts. Ensure the plot is labeled correctly.
• Describe the pattern of the forecasts, including the forecast intervals. Do they look reasonable?
• Using the “Pre-Covid” dataset, produce forecasts for the “Covid” time period using the ets and forecast functions in R. (Let R select the ETS model automatically).
• Compare the 2 sets of forecasts and comment.
• Quantify and compare the forecasted loss in revenue passenger miles arising from the 2 models.
• Provide a brief summary of what was done and a report of the forecasted loss in revenue passenger miles.
Add
