Name:______Stu ID:______Oct. 4, 2012

MIDTERM EXAMINATION Statistics 153 D. R. Brillinger

The exam is closed book. There are 2 questions, and a bonus question.. Answer both. Answer in the space provided. Show your work. Gfor your answers.

A calculator may be used: but please use only the following analytical keys: add, subtract, multiply, divide, square root.

You have exactly 50 minutes for the exam.

Please no questions to the proctor. If you are not sure of the meaning of a question, set down an interpretation, and provide your answer.

Question 1. (Chapter 4) Consider the series

Yt = et + q et-4 + q et-8 + ... (*)

for t=... with {et} a zero mean white noise.

a) What general time series class is this an example of?

Pages 1, 2

b) Derive and discuss its autocovariance function.

c) Can this series be stationary? If yes, when?

d) Can it be represented as a stationary AR? If yes which one?

e) What is an invertible series/process?

f) Is (^) ever one? When?

g) Define the autocorrelation function, rho_h.

h) What are rho_0, rho_1, rho_2, rho_3, rho_4 for the series (*)?

i) What can you say about the partial correlation function of (*)?

Pages 3,4

Question 2. Consider the time series

Y_t = a cos wt + b sin w + X_t (**)

with X_t ... and a, b, w fixed and known.

a) Show Y_t = a cos wt + b sin w satisfies

Y_t - 2 cos wt Y_{t-1} + Y_{t-2} = 0

[Reminder. 2 cos phi cos psi = ]

b) Consider the time series

Y_t = a cos wt + b sin w + X_t (**)

with X_t ... and w fixed and known.

How would you use the "operator" PSI of

PSI Y_t = Y_t - 2 cos wt Y_{t-1} + Y_{t-2 (***)

to create a stationary series from (**)

Pages 5,6

Question 3.

Consider the series {u_t}, {v_t} defined by

u_t = phi u_{t-1} + e_t

v_t = phi v_{t-1{ + f_t

with |phi| < 1,where E and f are white noises with means 0 and variances sigma_e and sigma_f respectively, further all the e_ are independent of all the f_s for s,r = ...

Consider the series Y_t = u_t +V_t, and Z_t = U_t - V_t

a) Derive the time series distributions of Y_t and Z_t.

Question 4. {Seasonal]

How can the model be written as a linear model?

Buyes-Balot

How would you compute estimates of the seasonal effects?

ARIMA

Question ? a). Derive the acf of the process defined by

=

Sketch it.

Comment on the result.

[Seasonal] Monthly daya. What does the acf look like? Reason for your answer?

Question ? a). Derive the acf of the process defined by

Sketch it.

Comment on the result.

Chapter 3. Trends

Chapter 5. Models for nonstationary series

Chapter 6 Model specification

Check Chatfield, Harvey, Brockwell-Davis, Shumway-Stoffer