You are on page 1of 2

CSC425 Time series analysis and forecasting

Homework 2
Due by Friday October 2nd, 2015 before 11:59pm
Total points: 20
Reading assignment
o Chapter 2 sections 2.1 through 2.4
o Review course documents posted under week 3
For R users
o Review notes on R posted under R resources in the Documents page on the course
website.
o Useful resources are at http://www.statmethods.net/ and
http://www.ats.ucla.edu/stat/R/
PROBLEMS
Problem 1 [4pts]
Consider the following AR(2) time series process: rt 0 . 5 0 . 16 rt 2 a t , where {at} is a
Gaussian white noise series with mean zero and constant variance 2=0.02. Note that the AR(2)
process has coefficient 1=0 for rt-1 (Hint: All the information you need to solve this questions is
in the book in sections 2.4.1 and 2.4.4 and in my slides)
a) What is the mean of the time series rt?
b) Determine if the AR(2) model is stationary. Explain.
c) Assume that r100= - 0.01 and r99 = 0.02. Compute the 1-step and 2-step ahead forecasts
of the AR(2) series at the forecast origin t=100.
d) Compute the lag-1 and lag-2 autocorrelations of rt (expressions in Week 3 slides)
Problem 2 [14 points]
Consider the oranges.csv data set you analyzed in homework 1. The dataset contains monthly
prices in US dollars for metric tons of oranges. You already analyzed the serial correlation of the
time series, now you will be asked to build an AR model to predict orange sales.
a) Analyze the ACF and the PACF plots for the orange sales data and discuss which order p
model is suggested by the plots?
b) Fit an adequate AR model:
Examine the significance of the model coefficients, and discuss if all coefficients are
significantly different from zero.
Perform a residual analysis and discuss if the selected model is adequate
i. compute ACF functions of residuals,
ii. test if residuals are White Noise,
iii. plot histogram and normal quantile plots of residuals.
Discuss the results of your residual analysis, and draw conclusions on whether the
selected model is appropriate to describe the orange sales.
c) Write down the expression of the estimated Ar(p) model, and test if the AR model
represents a stationary process. Explain your results.

d) Use the selected AR model to compute up to 5 step-ahead forecasts starting from the last
observation in the dataset. Write down the forecasts and their margin of errors or
prediction intervals.
e) Discuss where the forecasts converge to after a few steps
Reflection Problem [2 pts]
Post a message on the discussion board reflecting on the topics in week 4. Indicate the
assignment in this module you found to be the easiest, the one you found to be the hardest, and
why. I created a new Thread called hw2 Comments.
Submission instructions
Submit the homework at the Course Web page http:d2l.depaul.edu in the Dropbox page. Keep a
copy of all your submissions!
1. Submit your answers in a word or pdf document. Make sure to explain in detail your
analyses, and include relevant output and graphs. You should attach code to your
submission or past code in your word document.
2. If you have questions about the homework, email me BEFORE the deadline.
3. The assignment will lose 10% of the points per day, after the due date.
4. Assignments submitted five days after the deadline will not be accepted.

You might also like