Scribd
Upload
134 views15 pages

Module 2 - ARIMA PDF

1. The document discusses univariate time series forecasting techniques, specifically autoregressive (AR), moving average (MA), and autoregressive integrated moving average (ARIMA) models. 2. The Box-Jenkins procedure for identifying and estimating ARIMA models is presented, which involves identifying the model structure, estimating parameters, diagnostic checking, and forecasting. 3. Seasonal ARIMA models are also discussed to account for seasonality, with examples provided of possible SARIMA structures.

Uploaded by

Sruti Suhasaria
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
134 views15 pages

Module 2 - ARIMA PDF

1. The document discusses univariate time series forecasting techniques, specifically autoregressive (AR), moving average (MA), and autoregressive integrated moving average (ARIMA) models. 2. The Box-Jenkins procedure for identifying and estimating ARIMA models is presented, which involves identifying the model structure, estimating parameters, diagnostic checking, and forecasting. 3. Seasonal ARIMA models are also discussed to account for seasonality, with examples provided of possible SARIMA structures.

Uploaded by

Sruti Suhasaria
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/ 15

Module 2

Univariate Time Series Models

Muhammed Rafi OPC


Assistant Professor
BASE University
Univariate forecasting technique
Why Forecasting ?
Autoregressive Integrated Moving Average
(ARIMA) model

Box & Jenkins - 1976

Why univariate time-series?


An appropriate economic theory to explain the
relationship between series may not be available

Data unavailability- use lagged dependent variable


to forecast the future values 2
Autoregressive Model
yt depends only on its own past values yt 1, yt 2 , yt 3 ,...

AR (1) yt  1 yt 1  ut where ut  wn(0,  2 )

AR (2) yt  1 yt 1   2 yt 2  ut

AR (p) yt  1 yt 1   2 yt 2  .............   p yt  p  ut
p
y t    i y t i  u t
i 1
3
yt  1 yt 1  ut yt 1  1 yt  ut 1
Moving Average Model
yt depends only on past error terms ut 1, ut 2 , ut 3 ,...

MA (1) yt  1ut 1  ut

MA (2) yt  1ut 1   2ut 2  ut

q
MA (q) y t    i u t i  u t
i 1
Autoregressive Moving Average Model

ARMA (1, 1) yt  1 yt 1  φ1ut 1  ut

ARMA (2, 1) yt  1 yt 1   2 yt 2  φ1ut 1  ut

ARMA (2, 2) yt  1 yt 1   2 yt 2  φ1ut 1  φ2ut 2  ut

p q
ARMA (p, q) y t    i y t i   φ j u t  j  u t
i 1 j 1

5
Autoregressive Integrated Moving Average (ARIMA) Model

Integrated Process and ARIMA Models


yt
• ARMA model can be fit for only if it is a stationary process.

If y t is an integrated series of order (1) then ARIMA (1,


1, 1) model can be defined as
yt  1yt 1  φut 1  ut

• Hence, ARIMA (p, d, q)


p q
 y     y
d d
  φ u u
t i t i j t j t
i 1 j 1
6
Box-Jenkins procedure
Steps in ARIMA Model-building.
Identify the
Model

Estimate
Parameters

Diagnostic bad Revise the Model


Checking
good

Forecast
7
1. Identifying the structure of ARIMA
ARIMA (p, d, q)
p q
 y     y
d d
  φ u u
t i t i j t j t
i 1 j 1
• d?
• p? Partial Autocorrelation Function (PACF)

• q? Autocorrelation Function (ACF)

8
1. Identifying the structure of ARIMA
Correlation coefficient
T

 (Y t  Y )( X t  X )
 t  K 1
T T

 (Y
t 1
t Y ) 2
(X
t 1
t  X )2

Autocorrelation coefficient
T

 (Y t  Yt )(Yt  k  Yt  k )
 t  K 1
T T

 (Y
t 1
t  Yt ) 2
 (Y
t 1
t k  Yt  k ) 2

Cov (Yt , Yt  k )

Var (Yt )Var (Yt  k )

Cov (Yt , Yt  k )

Var (Yt )
9
Direct effect
Yt Yt-2
Yt-1

Indirect effect

Partial autocorrelation coefficient


Yt   0  1Yt 1   2Yt 2 .......   k Yt k   t

Partial autocorrelation between yt and yt  k is  k

10
2. Estimate Parameters
• AR terms
• MA terms

3. Diagnostic checking
• Highest R-square
• Highest Log likelihood
• Lowest AIC/SC/HQC
• Lowest RMSE

4. Forecast

11
Seasonal ARIMA (SARIMA) model
• seasonality
ARIMA (p, d, q) (P, D, Q)m
p = Autoregressive term (AR)

q = Moving Average term (MA)


d = differencing
P = Seasonal Autoregressive term (SAR)
Q = Seasonal Moving Average term (SMA)
D = Seasonal differencing
yt  yt 1 yt  yt m yt  yt 4 yt  yt 12
m = The number of time steps for a single seasonal period.
12
ARIMA (p, d, q) (P, D, Q)m

ARIMA (1, 0, 0) (1, 0, 0)4

ARIMA (0, 0, 1) (0, 0, 1)4

ARIMA (1, 0, 1) (2, 0, 1)4

ARIMA (1, 0, 1) (1, 0, 2)12

ARIMA (1, 0, 1) (1, 0, 2)7

13
Information Criteria
Akaike Information
Criterion

T= No.of observations
k= No.of parameters

Schwarz bayesian
criterion

Hannan Quinn Information


Criterion

14
Thank You

15

You might also like