1   Assessment task

This assignment is assessable. It is worth 20% of your final mark for the course. Any questions you have about this assignment must be asked on the Ed discussion forum.

Your answers need to be submitted using the Canvas quiz for assignment 2. The R script used to complete this assignment must be uploaded using this Canvas quiz. This document must be your own work.

The assignment asks you to identify the values of P and Q for the causal and invertibleARMA(P, Q)

model that best characterises the autocorrelation structure of 5 different data series.

When identifying the ARMA models, start with the ACF AND PACF for the raw data.  As required, also consider the statistical significance of AR and MA coefficients when you fit ARMA(P, Q) models. Finally analyse the residuals of the fitted model, making sure they have minimal remaining serial correlation.

If you choose to use AIC or BIC information criteria to select your model, be very careful not to over-parameterise the model. These criteria, and automated model selection strategies based upon them have a strong tendency to overstate P and/or Q. Assignments responses that make use of automated model selection methods tend to get low marks.

This week the all of the datasets (there are 300 of them) are files in the data zip file. Each of the data sets is an xts variable contained in its own R data file (a file with an ‘ .RData” extension). Your assignment involves analysing your assigned dataset. The number of the data set that you are to analyse is listed beside your name in the tutorial resources page on Canvas.

Your dataset will include 5 time-series realisations, named:

● series1

● series2

● series3

● series4

● series5

Download the zip file containing all of the data files and extract its contents into the folder where you will do your work.

Make sure thatfolder containing the data files is your working directory in R Studio.