Scribd Logo
Upload

Welcome to Scribd!

  • 18 views

    Time Series Basics: Fin250f: Lecture 8.1 Spring 2010 Reading: Brooks, Chapter 5.1-5.7

    Uploaded by

    Philip Semple
    Time series models like autoregressive (AR), moving average (MA), and autoregressive moving average (ARMA) models are commonly used to model temporal dependence in economic and engineering data. These models make it possible to forecast future values from past values. Identification of the appropriate model structure involves examining autocorrelation functions (ACF) and partial autocorrelation functions (PACF) of the time series, as well as using information criteria that penalize model complexity. Parameter estimation is typically done with ordinary least squares for AR models and numerical maximum likelihood methods for MA and ARMA models. The fitted models are then evaluated using residual diagnostic checks to validate that the residuals exhibit white noise properties.

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PPT, PDF, TXT or read online from Scribd

    Time Series Basics: Fin250f: Lecture 8.1 Spring 2010 Reading: Brooks, Chapter 5.1-5.7

    Uploaded by

    Philip Semple
    18 views44 pages
    Time series models like autoregressive (AR), moving average (MA), and autoregressive moving average (ARMA) models are commonly used to model temporal dependence in economic and engineering data. These models make it possible to forecast future values from past values. Identification of the appropriate model structure involves examining autocorrelation functions (ACF) and partial autocorrelation functions (PACF) of the time series, as well as using information criteria that penalize model complexity. Parameter estimation is typically done with ordinary least squares for AR models and numerical maximum likelihood methods for MA and ARMA models. The fitted models are then evaluated using residual diagnostic checks to validate that the residuals exhibit white noise properties.

    Original Title

    tsbasics (2)

    Copyright

    © Attribution Non-Commercial (BY-NC)

    Available Formats

    PPT, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Time series models like autoregressive (AR), moving average (MA), and autoregressive moving average (ARMA) models are commonly used to model temporal dependence in economic and engineering data. These models make it possible to forecast future values from past values. Identification of the appropriate model structure involves examining autocorrelation functions (ACF) and partial autocorrelation functions (PACF) of the time series, as well as using information criteria that penalize model complexity. Parameter estimation is typically done with ordinary least squares for AR models and numerical maximum likelihood methods for MA and ARMA models. The fitted models are then evaluated using residual diagnostic checks to validate that the residuals exhibit white noise properties.

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PPT, PDF, TXT or read online from Scribd
    Download as ppt, pdf, or txt
    18 views44 pages

    Time Series Basics: Fin250f: Lecture 8.1 Spring 2010 Reading: Brooks, Chapter 5.1-5.7

    Uploaded by

    Philip Semple
    Time series models like autoregressive (AR), moving average (MA), and autoregressive moving average (ARMA) models are commonly used to model temporal dependence in economic and engineering data. These models make it possible to forecast future values from past values. Identification of the appropriate model structure involves examining autocorrelation functions (ACF) and partial autocorrelation functions (PACF) of the time series, as well as using information criteria that penalize model complexity. Parameter estimation is typically done with ordinary least squares for AR models and numerical maximum likelihood methods for MA and ARMA models. The fitted models are then evaluated using residual diagnostic checks to validate that the residuals exhibit white noise properties.

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PPT, PDF, TXT or read online from Scribd
    Download as ppt, pdf, or txt
    You are on page 1of 44

    Time Series Basics

    Fin250f: Lecture 8.1 Spring 2010 Reading: Brooks, chapter 5.1-5.7

    Outline
    Linear

    stochastic processes Autoregressive process Moving average process Lag operator Model identification

    PACF/ACF Information Criteria

    Stochastic Processes
    Yt (y1 , y2 , y3 ,K yT ) E(Yt | yt 1 , 2 , ) = ( | 1)

    Time Series Definitions


    Strictly

    stationary Covariance stationary Uncorrelated White noise

    Strictly Stationary
    All

    distributional features are independent of time


    F(yt , yt 1 , ) ( | 1 ,, )

    Weak or Covariance Stationary


    Variances

    and covariances independent

    of time
    E(yt ) =

    ( )( ) = 2 < ( )( + ) =

    Autocorrelation
    gj gj
    2

    j =

    g0 s - 1 t j 1

    White Noise
    E(yt ) =

    ( )( ) = 2 < ( )( + ) = 0 > 0

    White Noise in Words


    Weakly

    stationary All autocovariances are zero Not necessarily independent

    Time Series Estimates


    1 = ( )( + ) =1 = 0 : (0,1 / )

    Ljung-Box Statistic
    2 Q * = ( + 2) =1

    Linear Stochastic Processes


    Linear

    models Time series dependence Common econometric frameworks Engineering background

    Autoregressive Process, Order 1:AR(1)


    yt = + 1 + | |< 1
    2 ( ) =

    AR(1) Properties
    E(yt ) = + ( 1 ) = + ( ) () = 1 ( +1 ) = +

    More AR(1) Properties


    =0 yt = f yt - 1 + ut
    E(yt2 ) = E(f yt - 1 + ut )(f yt - 1 + ut ) E(yt2 ) = E(f 2 yt2- 1 ) + 2E(utf yt - 1 ) + s E(yt2 ) = E(f 2 yt2 ) + s
    2 u 2 su E(yt2 ) = = var(yt ) 2 (1 - f ) 2 u

    More AR(1) properties


    =0 yt = f yt - 1 + ut E(yt yt - 1 ) = E(f yt - 1 + ut )(yt - 1 )
    g 1 = E(yt yt - 1 ) = f s E(yt yt - 1 ) t1= =f 2 sy t j =f
    j 2 y

    AR(1): Zero mean form


    yt +1 = + + +1 ( +1 ) = = 1 (yt+1 - ) =f (yt - )+ ut+1

    AR(m) (Order m)

    yt = + +
    =1

    Moving Average Process of Order 1, MA(1)


    yt = + 1 +

    MA(1) Properties
    yt = + 1 + () = ( +1 ) = +

    ( )2 = (1 + )(1 + )
    2 ( ) = (1 + 2 ) 2 1 = ( )( 1 ) = (1 + )(2 + 1 ) = 1 = ( , 1 ) = (1 + 2 ) = 0 2

    MA(m)
    yt = + +
    =1
    2 2 ( ) = (1 + 12 + + ) 2 , ) = ( + +11 + + ) (

    , ) = 0 (

    >

    Stationarity
    Process

    not exploding For AR(1) | |< 1 All finite MA's are stationary More complex beyond AR(1)

    AR(1)->MA(infinity)
    yt = 1 +

    1 = 2 + 1 = (2 + 1 ) + = 2 + 1 +
    2

    = + , | |< 1
    =0

    =
    =0

    Lag Operator (L)


    Lyt = 1

    = =

    Using the Lag Operator (Mean adjusted form)


    yt = ( 1 ) + = ( ) + (1 )( ) =

    An important feature for L


    yt = 1 + (1 ) = 1 = (1 )

    = = ()
    =0 =0
    1 = () (1 ) =0

    MA(1) -> AR(infinity)


    yt = + 1 + = (1 + ) 1 ( ) = (1 + )

    ( ) ( ) =
    =0

    MA->AR
    yt = ( ) ( ) +
    =1

    = (1) 1 ( ) +
    =1

    | |< 1

    AR's and MA's


    Can

    convert any stationary AR to an infinite MA Exponentially declining weights Can only convert "invertible" MA's to AR's Stationarity and invertibility:

    Easy for AR(1), MA(1) More difficult for larger models

    Combining AR and MA ARMA(p,q) (more later)


    yt = + + +
    =1 =1

    Modeling Procedures Box/Jenkins


    Identification

    Determine structure

    How many lags? AR, MA, ARMA?

    Tricky Estimate the parameters

    Estimation

    Residual

    diagnostics Next section: Forecast performance and evaluation

    Identification Tools
    Diagnostics

    ACF, Partial ACF Information criteria Forecast

    Autocorrelation
    1 = ( )( + ) =1 = 0 : (0,1 / ) 95% [ 1.96 (1 / ),1.96 (1 / )]

    Partial Autocorrelation
    Correlation

    between y(t) and y(t-k) after removing all smaller (<k) correlations Marginal forecast impact from t-k given all earlier information

    Partial Autocorrelation
    yt = + 1,1 1 +

    = + 1,2 1 + 2,2 2 + = + 1,3 1 + 2,3 2 + 3,3 3 + = [ 1,1 , 2,2 , 3,3 ,]

    For an AR(1)
    yt = + 1 +

    ( = = ) (1) = (> 1) = 0

    AR(1) (0.9)

    For an MA(1)
    yt = + 1 + (1) = 1 = 1+2 (> 1) = 0 = ()

    = (1) 1 ( ) +
    =1

    = ( , 2 , 3 , 4 ,)

    MA(1) (0.9)

    General Features
    Autoregressive

    Decaying ACF PACF drops to zero beyond model order(p)

    Moving

    average

    Decaying PACF ACF drops to zero beyond model order(q)

    Dont

    count on things looking so good

    Information Criteria
    Akaike,

    AIC Schwarz Bayesian criterion, SBIC Hannan-Quinn, HQIC Objective:


    Penalize model errors Penalize model complexity Simple/accurate models

    Information Criteria
    k=number of parameters 2 2 AIC = ) + ( 2) + ) = ( ( 2 2) + = ( ( )) (

    Estimation
    Autoregressive

    AR

    OLS Biased(-), but consistent, and approaches normal distribution for large T

    Moving

    average MA and ARMA

    Numerical estimation procedures Built into many packages

    Matlab econometrics toolbox

    Residual Diagnostics
    Get

    model residuals (forecast errors) Run this time series through various diagnostics

    ACF, PACF, Ljung/Box, plots

    Should

    be white noise (no structure)

    You might also like