All questions must be answered using R Markdown and submitted to Canvas. Use a 12 point font. Submit an HTML file.
You are working for a consulting company that has been given the task of modelling and forecasting US retail sales of beer, wine, and liquor stores. The client has previously used seasonal naive methods to forecast this data set but is now wondering if other methods would produce more accurate forecasts. The client is interested in forecasts for the period August 2020 to July 2021. You are asked to write a short research report (no more than 6 pages) on the forecasting and analysis of this data set.
In preparing your report you should consider the following:
The style of your report matters. Write a report for someone who has taken this course.
Use the package quantmod or excel to download the data from FRED. The data symbol is MRTSSM4453USN.
Using the data from January 2010 to July 2020, describe the trend and cyclical components. How has COVID-19 affected the data set? In doing the analysis compare the results from the classical decomposition with the results from the STL decomposition.
Which forecasting method (seasonal naïve, tslm, Holt-Winters, stlf() method=”naïve”, stlf() method=”ets”, ETS) fits best and why? What do the residuals look like for the best fitting model. Use the data from January 2010 to December of 2018 for training models and the data from January 2019 to July 2020 for testing models.
Using the best fitting model, produce forecasts for August 2020 to July 2021.