Lec12 – Time Series Analysis - ARIMA

Lec12 – Time Series Analysis - ARIMA - Lecture 12 Time...

Info iconThis preview shows pages 1–12. Sign up to view the full content.

View Full Document Right Arrow Icon
Lecture 12 – Time Series Analysis: Part III – ARIMA Model Dr. Qing He [email protected] Fall 2012 University at Buffalo CIE/IE 500 Transportation Analytics - Fall 2012 1
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
Static vs. Dynamic, and Stationary vs. Nonstationary & The trend regression model in Lec 11 is a static model since a value x t only depends on t and not x t-1 , x t-2 , . .. & The random walk (with drift) model is dynamic since computation of x t uses x t-1 , however this model was not stationary (recall its covariance function is given by γ(s, t) = min(s, t)σ 2 ). & This leads us to a large class of dynamic models which are stationary —the ARMA models. & Properties of the random walk model and the ARMA model are combined in the ( nonstationary ) AR I MA model CIE/IE 500 Transportation Analytics - Fall 2012 2
Background image of page 2
Autoregressive Model Example & Consider the following AR(1) process & where w t ~ iid N(0,1). We simulate the process in R. CIE/IE 500 Transportation Analytics - Fall 2012 3 > plot(arima.sim(list(order=c(1,0,0), ar=-.9), n=100)) t t t w x x + - = - 1 9 . 0
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
Definition of AR(p) Model & An autoregressive model of order p is of the form & where x t is stationary, φ 1 , φ 2 , . . . , φ p are constants (φ p ≠0), w t is Gaussion white noise with mean zero, variance σ 2 w . It has mean µ given by & Notice that x t is auto-regressed on x t-1 ,…, x t-p however these regressors have random components. CIE/IE 500 Transportation Analytics - Fall 2012 4 t p t p t t t w x x x x + + + + + = - - - φ α ... 2 2 1 1 p μ - - - - = ... 1 2 1 (assuming the denominator is nonzero)
Background image of page 4
The Autoregressive Operator & The general mean zero AR model & which is equivalent to & can be reformulated as & and even more concisely as CIE/IE 500 Transportation Analytics - Fall 2012 5 where Definition (Autoregressive Operator) The autoregressive operator is defined to be p p B B B B φ - - - - = ... 1 ) ( 2 2 1
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
AR(1) & Let’s investigate a general AR(1) process given by & Iterating backwards gives CIE/IE 500 Transportation Analytics - Fall 2012 6
Background image of page 6
AR(1)— MA( ) Representation & Provided |φ| < 1, we can represent the AR(1) model as & In this representation, we easily see & and the autocovariance function is computed as . .. CIE/IE 500 Transportation Analytics - Fall 2012 7 ± MA( )
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
AR(1)— ACF Computation & Autocovariance CIE/IE 500 Transportation Analytics - Fall 2012 8
Background image of page 8
AR(1)— ACF Computation cont. & ACF & Note that ρ(h) satisfies the recursion CIE/IE 500 Transportation Analytics - Fall 2012 9 (with the usual initial constraint ρ(0) = 1).
Background image of page 9

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
Causality & From the AR(1) model & we can solve for x t-1 as & If |φ| > 1, then the above expression gives a stationary representation of x t-1 in terms of future values of x t . This is useless since we don’t know the future! CIE/IE 500 Transportation Analytics - Fall 2012 10 Definition (Causality) When a process does not depend on the future, such as the AR(1) with |φ| < 1, the process is called causal . If |φ|>1, it is the explosive case of this example, so the process is stationary, but it is also future dependent, and not causal .
Background image of page 10
Moving Average Model— MA(q) & Definition (Moving average model— MA(q)) & The moving average model of order q is defined to be & The above model can be compactly written as & where θ(B) is the moving average operator.
Background image of page 11

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
Image of page 12
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}

Page1 / 59

Lec12 – Time Series Analysis - ARIMA - Lecture 12 Time...

This preview shows document pages 1 - 12. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online