Chapter 15. Time Series: Descriptive Analyses, Models, and Forecasting

Similar documents
Types of Exponential Smoothing Methods. Simple Exponential Smoothing. Simple Exponential Smoothing

Time series Decomposition method

Summer Term Albert-Ludwigs-Universität Freiburg Empirische Forschung und Okonometrie. Time Series Analysis

Smoothing. Backward smoother: At any give T, replace the observation yt by a combination of observations at & before T

Exponential Smoothing

Licenciatura de ADE y Licenciatura conjunta Derecho y ADE. Hoja de ejercicios 2 PARTE A

Wisconsin Unemployment Rate Forecast Revisited

Nature Neuroscience: doi: /nn Supplementary Figure 1. Spike-count autocorrelations in time.

OBJECTIVES OF TIME SERIES ANALYSIS

Forecasting. Summary. Sample StatFolio: tsforecast.sgp. STATGRAPHICS Centurion Rev. 9/16/2013

Diebold, Chapter 7. Francis X. Diebold, Elements of Forecasting, 4th Edition (Mason, Ohio: Cengage Learning, 2006). Chapter 7. Characterizing Cycles

Distribution of Least Squares

Estimation Uncertainty

Financial Econometrics Jeffrey R. Russell Midterm Winter 2009 SOLUTIONS

Lecture 15. Dummy variables, continued

Solutions to Exercises in Chapter 12

Lecture 3: Exponential Smoothing

Solutions to Odd Number Exercises in Chapter 6

Vehicle Arrival Models : Headway

Navneet Saini, Mayank Goyal, Vishal Bansal (2013); Term Project AML310; Indian Institute of Technology Delhi

Modeling and Forecasting Volatility Autoregressive Conditional Heteroskedasticity Models. Economic Forecasting Anthony Tay Slide 1

EVALUATING FORECASTING MODELS FOR UNEMPLOYMENT RATES BY GENDER IN SELECTED EUROPEAN COUNTRIES

PROC NLP Approach for Optimal Exponential Smoothing Srihari Jaganathan, Cognizant Technology Solutions, Newbury Park, CA.

14 Autoregressive Moving Average Models

GDP Advance Estimate, 2016Q4

Nonstationarity-Integrated Models. Time Series Analysis Dr. Sevtap Kestel 1

Forecasting optimally

STAD57 Time Series Analysis. Lecture 17

STAD57 Time Series Analysis. Lecture 17

Econ Autocorrelation. Sanjaya DeSilva

b denotes trend at time point t and it is sum of two

Section 4 NABE ASTEF 232

Quarterly ice cream sales are high each summer, and the series tends to repeat itself each year, so that the seasonal period is 4.

Forward guidance. Fed funds target during /15/2017

Stationary Time Series

TIME SERIES ANALYSIS. Page# 1

Distribution of Estimates

3.1 More on model selection

Properties of Autocorrelated Processes Economics 30331

Cointegration and Implications for Forecasting

ACE 562 Fall Lecture 8: The Simple Linear Regression Model: R 2, Reporting the Results and Prediction. by Professor Scott H.

ACE 564 Spring Lecture 7. Extensions of The Multiple Regression Model: Dummy Independent Variables. by Professor Scott H.

ECON 482 / WH Hong Time Series Data Analysis 1. The Nature of Time Series Data. Example of time series data (inflation and unemployment rates)

4.1 Other Interpretations of Ridge Regression

FORECASTING WITH REGRESSION

Institutional Assessment Report Texas Southern University College of Pharmacy and Health Sciences "P1-Aggregate Analyses of 6 cohorts ( )

References are appeared in the last slide. Last update: (1393/08/19)

Wednesday, November 7 Handout: Heteroskedasticity

- The whole joint distribution is independent of the date at which it is measured and depends only on the lag.

Regression with Time Series Data

CHAPTER 9. Exercise Solutions

Linear Combinations of Volatility Forecasts for the WIG20 and Polish Exchange Rates

Outline. lse-logo. Outline. Outline. 1 Wald Test. 2 The Likelihood Ratio Test. 3 Lagrange Multiplier Tests

Reliability of Technical Systems

Kriging Models Predicting Atrazine Concentrations in Surface Water Draining Agricultural Watersheds

Linear Gaussian State Space Models

Nonlinearity Test on Time Series Data

The Simple Linear Regression Model: Reporting the Results and Choosing the Functional Form

Uncertainty in predictive modelling

Vectorautoregressive Model and Cointegration Analysis. Time Series Analysis Dr. Sevtap Kestel 1

NCSS Statistical Software. , contains a periodic (cyclic) component. A natural model of the periodic component would be

Variance Bounds Tests for the Hypothesis of Efficient Stock Market

Article from. Predictive Analytics and Futurism. July 2016 Issue 13

THE UNIVERSITY OF TEXAS AT AUSTIN McCombs School of Business

I. Return Calculations (20 pts, 4 points each)

Use of Unobserved Components Model for Forecasting Non-stationary Time Series: A Case of Annual National Coconut Production in Sri Lanka

Methodology. -ratios are biased and that the appropriate critical values have to be increased by an amount. that depends on the sample size.

On Measuring Pro-Poor Growth. 1. On Various Ways of Measuring Pro-Poor Growth: A Short Review of the Literature

STA 114: Statistics. Notes 2. Statistical Models and the Likelihood Function

Introduction D P. r = constant discount rate, g = Gordon Model (1962): constant dividend growth rate.

Intermittent Demand Forecast and Inventory Reduction Using Bayesian ARIMA Approach

Chapter 16. Regression with Time Series Data

T L. t=1. Proof of Lemma 1. Using the marginal cost accounting in Equation(4) and standard arguments. t )+Π RB. t )+K 1(Q RB

1.6. Slopes of Tangents and Instantaneous Rate of Change

Answers to Exercises in Chapter 7 - Correlation Functions

Fourier Transformation on Model Fitting for Pakistan Inflation Rate

R t. C t P t. + u t. C t = αp t + βr t + v t. + β + w t

Chapter 11. Heteroskedasticity The Nature of Heteroskedasticity. In Chapter 3 we introduced the linear model (11.1.1)

ACE 562 Fall Lecture 4: Simple Linear Regression Model: Specification and Estimation. by Professor Scott H. Irwin

( ) = b n ( t) n " (2.111) or a system with many states to be considered, solving these equations isn t. = k U I ( t,t 0 )! ( t 0 ) (2.

Chapter 7 Response of First-order RL and RC Circuits

Volatility. Many economic series, and most financial series, display conditional volatility

In this paper the innovations state space models (ETS) are used in series with:

Exponential Weighted Moving Average (EWMA) Chart Under The Assumption of Moderateness And Its 3 Control Limits

Frequency independent automatic input variable selection for neural networks for forecasting

04. Kinetics of a second order reaction

Hypothesis Testing in the Classical Normal Linear Regression Model. 1. Components of Hypothesis Tests

Innova Junior College H2 Mathematics JC2 Preliminary Examinations Paper 2 Solutions 0 (*)

Chapter 5. Heterocedastic Models. Introduction to time series (2008) 1

An Initial Study on the Forecast Model for Unemployment Rate. Mohd Nadzri Mohd Nasir, Kon Mee Hwa and Huzaifah Mohammad 1

Dynamic Econometric Models: Y t = + 0 X t + 1 X t X t k X t-k + e t. A. Autoregressive Model:

Econ107 Applied Econometrics Topic 7: Multicollinearity (Studenmund, Chapter 8)

The Implementation of Business Decision-Making Tools in Incident Rate Prediction

Section 7.4 Modeling Changing Amplitude and Midline

Has the Business Cycle Changed? Evidence and Explanations. Appendix

Course Notes for EE227C (Spring 2018): Convex Optimization and Approximation

A Dynamic Model of Economic Fluctuations

Dynamic Models, Autocorrelation and Forecasting

Wednesday, December 5 Handout: Panel Data and Unobservable Variables

LabQuest 24. Capacitors

Transcription:

Chaper 15 Time Series: Descripive Analyses, Models, and Forecasing

Descripive Analysis: Index Numbers Index Number a number ha measures he change in a variable over ime relaive o he value of he variable during a specific base period Simple Index Number index based on he relaive changes (over ime) in he price or quaniy of a single commodiy In d e x n u m b e r a im e T im e s e r ie s v a lu e a im e T im e s e r ie s v a lu e a b a s e p e r io d 100 I Y Y i 0 100

Descripive Analysis: Index Numbers Laspeyres and Paasche Indexes compared The Laspeyres Index weighs by he purchase quaniies of he baseline period The Paasche Index weighs by he purchase quaniies of he period he index value represens. Laspeyres Index is mos appropriae when baseline purchase quaniies are reasonable approximaions of purchases in subsequen periods. Paasche Index is mos appropriae when you wan o compare curren o baseline prices a curren purchase levels

Descripive Analysis: Index Numbers Calculaing a Laspeyres Index Collec price info for he k price series o be used, denoe as P 1, P 2 P k Selec a base period 0 Collec purchase quaniy info for base period, denoe as Q 10, Q 20..Q k0 Calculae weighed oals for each ime period using he k formula Q P i 0 i k i 1 Q P i 0 i i 1 Calculae he index using he formula I 100 k Q P i 0 i 0 i 1

Descripive Analysis: Index Numbers Calculaing a Paasche Index Collec price info for he k price series o be used, denoe as P 1, P 2 P k Selec a base period 0 Collec purchase quaniy info for every period, denoe as Q 1, Q 2..Q k Calculae he index for ime using he formula I k i 1 k i 1 Q Q i i P P i i 0 100

Descripive Analysis: Exponenial Smoohing Exponenial smoohing is a ype of weighed average, ha applies a weigh w o pas and curren values of he ime series. Exponenial smoohing consan (w) lies beween 0 and 1, and smoohed series E is calculaed as: E Y 1 1 E w Y (1 w ) E 2 2 1 E w Y (1 w ) E 3 3 2 E w Y (1 w ) E 1

Descripive Analysis: Exponenial Smoohing Selecion of smoohing consan w is made by researcher. Small values of w give less weigh o curren value, yield a smooher series Large values of w give more weigh o curren value, yield a more variable series

Time Series Componens 4 componens of ime series models: T secular or long-erm rend C cyclical rend S seasonal effec R residual effec These 4 componens are from a widely used model called he addiive model: Y = T + C + S + R

Forecasing: Exponenial Smoohing Calculaion of Exponenially Smoohed Forecass Calculae exponenially smoohed values E 1, E 2, E for observed ime series Y 1, Y 2, Y. Used las smoohed value o forecas he nex ime series value F E F w Y F 1 Assuming ha Y is relaively free of rend and seasonal componens, use he same forecas for all fuure values of Y : F F F 2 1 F 3 1

Forecasing Trends: The Hol- Winers Model The Hol-Winers model adds a rend componen o he forecas. Calculaing Componens of he Hol-Winers Model 1. Selec exponenial smoohing consan w 2. Selec rend smoohing consan v 3. Calculae he wo componens E and T from ime series Y beginning a ime =2 E 2 =Y 2 T 2 =Y 2 -Y 1 E 3 =wy 3 +(1-w)(E 2 +T 2 ) T 3 =v(e 3 -E 2 )+(1-v)T 2 E =wy +(1-w)(E -1 +T -1 ) T =v(e -E -1 )+(1-v)T -1

Forecasing Trends: The Hol- Winers Model Hol-Winers Forecasing 1. Calculae he exponenially smoohed and rend componens E and T for each observed value of Y ( >2) 2. Calculae he one-sep-ahead forecas using F +1 =E +T 3. Calculae he k-sep-ahead forecas using F +k =E +kt

Measuring Forecas Accuracy: MAD and RMSE Measures of Forecas Accuracy for m Forecass Mean absolue Deviaion (MAD) M A D Mean absolue percenage error (MAPE) Roo mean squared error (RMSE) R M S E m 1 Y m M A P E m 1 m F 1 Y m Y Y F 2 m F 100

Forecasing Trends: Simple Linear Regression Simple Linear Regression is he simples inferenial forecasing model Afer fiing he regression line o exising daa, he leas squares model can be used o forecas fuure values of he dependen variable Two problems are associaed wih using a LSM o forecas ime series: 1. Forecasing falls ouside of he experimenal region, increases widh of predicion inervals 2. Cyclical effecs are no buil in o he model, and inroduce he problem of correlaed error

Seasonal Regression Models Use of muliple regression model wih dummy variables o describe seasonal differences is common In he following example, dummy variables are se up for Quarers 1, 2 and 3 Model ha reflecs seasonal componen and expeced growh in usage is: E ( Y ) Q Q Q 0 1 2 1 3 2 4 3

Seasonal Regression Models Daa o be used is in he following able: Quarerly Power Loads (megawas) for a Souhern Uiliy Company, 1992-2003 Year Quarer Power Load Year Quarer Power Load 1992 1 68.8 1998 1 130.6 2 65.0 2 116.8 3 88.4 3 144.2 4 69.0 4 123.3 1993 1 83.6 1999 1 142.3 2 69.7 2 124.0 3 90.2 3 146.1 4 72.5 4 135.5 1994 1 106.8 2000 1 147.1 2 89.2 2 119.3 3 110.7 3 138.2 4 91.7 4 127.6 1995 1 108.6 2001 1 143.4 2 98.9 2 134.0 3 120.1 3 159.6 4 102.1 4 135.1 1996 1 113.1 2002 1 149.5 2 94.2 2 123.3 3 120.5 3 154.4 4 107.4 4 139.4 1997 1 116.2 2003 1 151.6 2 104.4 2 133.7 3 131.7 3 154.5 4 117.9 4 135.1

Seasonal Regression Models Resul of regression analysis is:

Seasonal Regression Models Forecas resuls and acual values for 2004 are:

Seasonal Regression Models Use of muliplicaive models ofen provides a beer forecasing model when he ime series is changing a an increasing rae over ime. Muliplicaive model for Power Load problem would be: ln ( ) E Y Q Q Q 0 1 2 1 3 2 4 3 Taking anilogarihm of boh sides shows muliplicaive naure: Y e x p e x p e x p Q Q Q e x p 0 1 2 1 3 2 4 3 Consan Secular Seasonal Componen Residual rend Componen

Auocorrelaion and he Durbin- Wason Tes A residual paern as illusraed here suggess ha auocorrelaion may be an issue. Auocorrelaion is he correlaion beween ime series residuals a differen poins in ime. Correlaion beween neighboring residuals is called firs-order auocorrelaion

Auocorrelaion and he Durbin- Wason Tes Durbin-Wason d saisic is calculaed o es for he presence of firs-order auocorrelaion n 2 R R 1 2 d R a n g e o f d : 0 d 4 n 2 R 1 If residuals are uncorrelaed, hen d 2. If residuals are posiively auocorrelaed, hen d<2, and if auocorrelaion is very srong, d 0. If residuals are negaively auocorrelaed, hen d>2, and if auocorrelaion is very srong, d 4.

Auocorrelaion and he Durbin- Wason Tes One-Tailed Tes H 0 :No firs-order auocorrelaion H a :Posiive firs-order auocorrelaion (or H a :Negaive firs-order auocorrelaion) Two-Tailed Tes H 0 :No firs-order auocorrelaion H a :Posiive or negaive firs-order auocorrelaion Tes Saisic d n 2 R n 1 R R 2 1 2 Rejecion region: d< d L, [or (4-d)< d L, if H a :Negaive firs-order auocorrelaion] Where d L, is he lower abled value corresponding o k independen variables and n observaions. The corresponding upper value d U, defines a possibly significan region beween d L, and d U, Rejecion region: d< d L,/2 or (4-d)< d L,/2 Where d L,/2 is he lower abled value corresponding o k independen variables and n observaions. The corresponding upper value d U,/2 defines a possibly significan region beween d L,/2 and d U,/2