Scribd
Upload
135 views7 pages

Backshift Notation: Nonseasonal Processes in Backshift Form

This document discusses backshift notation for representing nonseasonal ARIMA processes. It defines the backshift operator B and differencing operator (1-B). Examples are provided for converting ARIMA processes between algebraic and backshift forms, including ARIMA(1,1,1). A procedure is outlined for writing any nonseasonal ARIMA process in backshift notation. Exercises are included to practice converting additional processes between forms.

Uploaded by

Túlio Igor Soares Pereira
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
135 views7 pages

Backshift Notation: Nonseasonal Processes in Backshift Form

This document discusses backshift notation for representing nonseasonal ARIMA processes. It defines the backshift operator B and differencing operator (1-B). Examples are provided for converting ARIMA processes between algebraic and backshift forms, including ARIMA(1,1,1). A procedure is outlined for writing any nonseasonal ARIMA process in backshift notation. Exercises are included to practice converting additional processes between forms.

Uploaded by

Túlio Igor Soares Pereira
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

Econ 415 Dr. D.

Tran

CHAPTER 12

NONSEASONAL PROCESSES IN BACKSHIFT FORM

1. BACKSHIFT NOTATION
Backshift operator B, defined as

BYt = Yt-1
BZt = Zt-1

B2Yt = Yt-2

B3Zt = Zt-3
BC = C (C is a constant)

BkC = C (where k is a constant)

Since Zt = Yt ! where is a constant mean of Yt

BZt = B(Yt ! )
= BYt ! B

Zt!1 = Yt!1 !

BZt-1 = Zt-2

2. DIFFERENCING OPERATOR (1!B)


First Order Simple (regular) Differencing:

(1!B)Zt = Zt ! Zt!1

Second Order Simple (regular) Differencing:

(1!B)2Zt = (1!B)(1!B)Zt

= (1 ! 2B + B2)Zt

= (1!B)(Zt ! Zt!1)

= (Zt ! Zt-1) ! (1!B)Zt!1

= Zt ! Zt!1 ! Zt!1 + Zt!2


= Zt ! 2Zt!1 + Zt!2
Econ 415/Tran Chapter 12/Page 2

From now on, when we refer to a realization (sample) we write

Zt = Yt ! (where is the realization mean)

But when we refer to the process, we write

Zt = Yt ! (where is the mean of the process).

Thus Zt could be deviation from realization or process mean, depending on the context.

To simplify further we write difference operator as

= (1-B)Zt = Zt - Zt!1

= (1-B)2Zt

= Zt ! Zt!1 ! (Zt!1 ! Zt!2)

= Zt ! 2Zt!1 + Zt!2

In general,

3. NONSEASONAL ARIMA MODELS


Examples on backshift notations:

AR(1)
Algebraic Form:

(1)

Backshift Form:

(2)
Econ 415/Tran Chapter 12/Page 3

substituting , we get

Now let , then (2) is equivalent to (1).

MA(1)
Algebraic Form:

(3)

Backshift form:

(4)

Substituting , we get

Let C = then (3) and (4) are equivalent.

4. DIFFERENCING
Consider a process for a variable Yt which must be differenced once because its mean is not
constant. Suppose that the first difference of Yt are a series of independent random shocks, a
random walk model, i.e.
Econ 415/Tran Chapter 12/Page 4

Yt ! Yt!1 = at

To write this in backshift from, multiply Zt = Yt ! by (say gradient):

Zt ! Zt-1 = at

(Yt ! ) - (Yt-1 ! ) = at

Yt ! Yt-1 = at

Note that drops out when is applied to Zt = Yt ! . This happens with any process
even when we apply multi-level differencing, i.e. when d > 0.

Examples

ARIMA(1,1,1)

For ARIMA(1,0,1) we have . Now


to Yt:

Algebraic form:

First apply the first difference to Yt and Yt-1, then express it in a common algebraic form:

(1)

Backshift Form:

(2)

5. SUMMARY OF PROCEDURE
To write nonseasonal ARIMA processes in backshift form, follow the following
procedure:

1. Start with a variable Yt that has been transformed (if necessary) so that it has a constant
variance (by Box-Cox transform).

2. Write Yt in deviations from its mean, i.e. use Zt where Zt = Yt - .


Econ 415/Tran Chapter 12/Page 5

3. Multiply Zt by the differencing operator (1 - B)d or to ensure that we have a variable


whose mean is stationary.

4. Multiply the result from step 3 by the AR operator whose general form is

( ).

For a specific process, assign the appropriate numerical value to p, the order of the AR
part of the process. If any coefficient with subscripts less than p are zero, exclude
those terms from the AR operator.

5. Multiply the random shock at by the MA operator whose general form is

( ).

For specific process, assign the appropriate numerical value to q, the order of the MA
portion of the process. If any the coefficients at lag less than q are zero, exclude them
from the MA process.

6. Equate the results from step 4 and step 5.

7. Combining the above 6 steps, a nonseasonal process in backshift notation has this general
form:

In a more compact form, we write:

where

Now we want to show that both forms are equivalent for ARIMA(1,1,1).

Algebraic Form:

(1)

Backshift Form:
Econ 415/Tran Chapter 12/Page 6

(2)

Substituting Zt = Yt ! , we obtain

Thus let C = 0, then (1) and (2) are equivalent. This means that the first differences have a mean
of zero.

6. EXERCISES:

(1) Write the following in both algebraic form and backshift form:

(a) ARIMA (1,1,2)


(b) ARIMA (0,2,1)
(c) ARIMA (2,0,2)
(2) Write the following in back shift form:

(a)

(b)

(c)

(3) Write the following in both ARIMA(p,d,q) notation and common algebraic form:

(a)

(b)
Econ 415/Tran Chapter 12/Page 7

(c)

You might also like