Time Series Models for
B
i
Business
TSMB Lecture 3
1

3
.
Nonstationary Models
and Regression
In this chapter we will examine
the problem of finding
the problem of finding
an appropriate model for data
that does not seem to be
that does not seem to be
generated by
t ti
ti
i
a stationary time series.
2

Nonstationary Data
If the data
(i)
hibit
t d
i ti
f
t ti
it
(i) exhibit no apparent deviation from stationarity,
a
nd,
(ii) have a rapidly decreasing ACVF,
then, we attempt to fit an ARMA model
to the mean-corrected data
using the techniques of Chapter 2
using the techniques of Chapter 2.
If (i) and/or (ii) are not satisfied,
then
then,
differencing often achieves this,
leading us to consider the class of ARIMA models.
3