x 1 = x i1 x i2 y = x 1 β x K β K + ε, x i =

Similar documents
(X i X) 2. n 1 X X. s X. s 2 F (n 1),(m 1)

7. GENERALIZED LEAST SQUARES (GLS)

Questions and Answers on Heteroskedasticity, Autocorrelation and Generalized Least Squares

Ma 3/103: Lecture 24 Linear Regression I: Estimation

Heteroskedasticity and Autocorrelation

Introduction to Econometrics Midterm Examination Fall 2005 Answer Key

Econometrics of Panel Data

EC3062 ECONOMETRICS. THE MULTIPLE REGRESSION MODEL Consider T realisations of the regression equation. (1) y = β 0 + β 1 x β k x k + ε,

Formulary Applied Econometrics

FENG CHIA UNIVERSITY ECONOMETRICS I: HOMEWORK 4. Prof. Mei-Yuan Chen Spring 2008

Section 6: Heteroskedasticity and Serial Correlation

Econ 620. Matrix Differentiation. Let a and x are (k 1) vectors and A is an (k k) matrix. ) x. (a x) = a. x = a (x Ax) =(A + A (x Ax) x x =(A + A )

GLS and related issues

Topic 6: Non-Spherical Disturbances

Econometrics of Panel Data

Well-developed and understood properties

Ch.10 Autocorrelated Disturbances (June 15, 2016)

Econometrics. Week 4. Fall Institute of Economic Studies Faculty of Social Sciences Charles University in Prague

Problems. Suppose both models are fitted to the same data. Show that SS Res, A SS Res, B

Econometrics of Panel Data

General Linear Test of a General Linear Hypothesis. Copyright c 2012 Dan Nettleton (Iowa State University) Statistics / 35

3. For a given dataset and linear model, what do you think is true about least squares estimates? Is Ŷ always unique? Yes. Is ˆβ always unique? No.

Advanced Quantitative Methods: Regression diagnostics

9. AUTOCORRELATION. [1] Definition of Autocorrelation (AUTO) 1) Model: y t = x t β + ε t. We say that AUTO exists if cov(ε t,ε s ) 0, t s.

Clustering as a Design Problem

Working Paper Series. An Empirical Investigation of Direct and Iterated Multistep Conditional Forecasts. Michael W. McCracken and Joseph McGillicuddy

Non-Spherical Errors

Total Least Squares Approach in Regression Methods

Mathematical and Information Technologies, MIT-2016 Mathematical modeling

Weighted Least Squares

Linear Regression. In this problem sheet, we consider the problem of linear regression with p predictors and one intercept,

Econ 583 Final Exam Fall 2008

. a m1 a mn. a 1 a 2 a = a n

K e sub x e sub n s i sub o o f K.. w ich i sub s.. u ra to the power of m i sub fi ed.. a sub t to the power of a

Different types of regression: Linear, Lasso, Ridge, Elastic net, Ro

Lecture 4: Heteroskedasticity

Quick Review on Linear Multiple Regression

Lecture Notes on Different Aspects of Regression Analysis

Lecture 14 Simple Linear Regression

A COMPARISON OF HETEROSCEDASTICITY ROBUST STANDARD ERRORS AND NONPARAMETRIC GENERALIZED LEAST SQUARES

Lecture 1: OLS derivations and inference

G023: Econometrics.

Lecture 15. Hypothesis testing in the linear model

Nonparametric Regression. Badr Missaoui

Nonstationary Panels

Massachusetts Institute of Technology Department of Economics Time Series Lecture 6: Additional Results for VAR s

Multivariate Regression Analysis

Factor Model Risk Analysis

Improved Inference for First Order Autocorrelation using Likelihood Analysis

Introductory Econometrics

Ma 3/103: Lecture 25 Linear Regression II: Hypothesis Testing and ANOVA

Heteroskedasticity in Panel Data

Advanced Econometrics I

Heteroskedasticity in Panel Data

Linear Regression Model. Badr Missaoui

Time Series Analysis

STAT 213 Interactions in Two-Way ANOVA

Regression. ECO 312 Fall 2013 Chris Sims. January 12, 2014

Single Equation Linear GMM with Serially Correlated Moment Conditions

Generalized Least Squares

Auto-correlation of Error Terms

20.1. Balanced One-Way Classification Cell means parametrization: ε 1. ε I. + ˆɛ 2 ij =

ECON 3150/4150, Spring term Lecture 7

UNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS

Spatial Models in Econometrics: Section 13 1

Economics 672 Fall 2017 Tauchen. Jump Regression

GENERALISED LEAST SQUARES AND RELATED TOPICS

Financial Time Series Analysis Week 5

Single Equation Linear GMM with Serially Correlated Moment Conditions

POL 572 Multivariate Political Analysis

Econometrics. Andrés M. Alonso. Unit 1: Introduction: The regression model. Unit 2: Estimation principles. Unit 3: Hypothesis testing principles.

Applied Econometrics (MSc.) Lecture 3 Instrumental Variables

STAT Regression Methods

Vector Auto-Regressive Models

Basic Distributional Assumptions of the Linear Model: 1. The errors are unbiased: E[ε] = The errors are uncorrelated with common variance:

Intermediate Econometrics

Ch 2: Simple Linear Regression

Economics 582 Random Effects Estimation

VAR Models and Applications

Peter Hoff Linear and multilinear models April 3, GLS for multivariate regression 5. 3 Covariance estimation for the GLM 8

Understanding Regressions with Observations Collected at High Frequency over Long Span

Advanced Quantitative Methods: ordinary least squares

NBER WORKING PAPER SERIES ROBUST STANDARD ERRORS IN SMALL SAMPLES: SOME PRACTICAL ADVICE. Guido W. Imbens Michal Kolesar

Empirical Economic Research, Part II

Lecture 19 Multiple (Linear) Regression

Heteroskedasticity. y i = β 0 + β 1 x 1i + β 2 x 2i β k x ki + e i. where E(e i. ) σ 2, non-constant variance.

Econometrics of Panel Data

Introduction to Econometrics Final Examination Fall 2006 Answer Sheet

Specification errors in linear regression models

Økonomisk Kandidateksamen 2004 (I) Econometrics 2. Rettevejledning

Simple Linear Regression

Biostatistics 533 Classical Theory of Linear Models Spring 2007 Final Exam. Please choose ONE of the following options.

7 Day 3: Time Varying Parameter Models

Bootstrapping Heteroskedasticity Consistent Covariance Matrix Estimator

Chapter 6. Panel Data. Joan Llull. Quantitative Statistical Methods II Barcelona GSE

Summer School in Statistics for Astronomers V June 1 - June 6, Regression. Mosuk Chow Statistics Department Penn State University.

11 Hypothesis Testing

Robust Standard Errors in Small Samples: Some Practical Advice

In the bivariate regression model, the original parameterization is. Y i = β 1 + β 2 X2 + β 2 X2. + β 2 (X 2i X 2 ) + ε i (2)

Stationary Stochastic Time Series Models

Transcription:

x k T x k k = 1,, K T K X X 1 1 1 x 1 = 1 β 1 y T y 1 y T ε T T 1 x i1 x i2 y = x 1 β 1 + + x K β K + ε, x i = y T 1 = X T K β K 1 + ε T 1. x it T 1 y x 1 x K y = Xβ + ε X T K K E[ε i x j1, x j2,, x jk ] = 0 i ε i E[ε X] = 0. ε i σ 2 ε j E[εε X] = σ 2 I. (x j1, x j2,, x jk ) X ε X N(0, σ 2 I). β 1 β 2 β K ( ) 2 T T K RSS = ˆε ˆε = ˆε 2 t = y t x it β i t=1 t=1 i=1

ˆβ 1 ˆβ = ˆβ 2 = (X X) 1 X y. ˆβ k σ 2 s 2 = RSS T T K = t=1 ˆε2 t T K K K K T K ( ˆβ) = s 2 (X X) 1. ˆβ = (X X) 1 X y = β + (X X) 1 X ε s 2 T t=1 = ˆε2 t T K ( ˆβ) = s 2 (X X) 1. t ˆβ i SE( ˆβ t(t K) i ) SE( ˆβ i ) = ( ˆβ) ii F F β F F RRSS URSS URSS T K m F (m, T K) URSS RRSS m T K

RRSS URSS t F t 2 (T K) F (1, T K). (X X) x k x j r kj = s kj s k s j. t 1 V IF k = 1 Rk 2 R 2 k x k

t R 2 y

E[ε 2 i X] T T 1 T 2 s 2 1 = ˆε 1 ˆε 1/(T 1 K), s 2 2 = ˆε 2 ˆε 2/(T 2 K) GQ GQ = s2 1 s 2 2 s 2 1 > s 2 2 F (T 1 K, T 2 K) GQ y t = β 1 + β 2 x 2t + β 3 x 3t + ε t. (ε t ) = σ 2 ˆε t ˆε 2 t = α 1 + α 2 x 2t + α 3 x 3t + α 4 x 2 2t + α 5 x 2 3t + α 6 x 2t x 3t + ν t. (ε t ) = E[ε 2 t ] E[ε t ] = 0 ˆε 2 t F ˆε 2 t F R 2 R 2 R 2 R 2 T T R 2 χ 2 (m) m F

α 2 = α 3 = α 4 = α 5 = α 6 = 0 χ 2 x k σε 2 t = σεx 2 α kt y t = β 1 + β 2 x 2t + + β K x Kt + ε t ˆε 2 t ˆε 2 t = γ + α x kt + ν t α t t = ˆαˆσ ˆα α σ 2 σ 2 z t (ε t ) = σ 2 z 2 t. z t y t 1 x 2t x 3t = β 1 + β 2 + β 3 + ν t z t z t z t z t ν t = εt z t

se( ˆβ T 1 ) HC = t=1 (x t x) 2ˆε 2 t ( T ) 2. t=1 (x t x) 2 se( ˆβ T k ) HC = t=1 ˆω2 tk ˆε2 t ) 2 ( T t=1 ˆω2 tk ˆω tk 2 x k y t = β 1 +β 2 x 2t + +β K x Kt +ε t ˆε 2 t K 1 T (K 1) (ˆω tpk 2 ) se( ˆβ T k ) HC = t=1 ˆω2 tk ˆε2 t ) 2. ( T t=1 ˆω2 tk ˆε t ˆε t 1 ˆε t (ˆε t 1, ˆε t ) (ˆε t 1, ˆε t ) ˆε t

µ r ± 1.96σ r µ r σ r µ r = 2T 1T 2 2T 1 T 2 (2T 1 T 2 T 1 T 2 ) + 1, σ r = T 1 + T 2 (T 1 + T 2 ) 2 (T 1 + T 2 1) r T 1 T 2 T t ε t = ρε t 1 + ν t ν t N(0, σ 2 nu) H 0 : ρ = 0, H 1 : ρ 0. T t=2 DW = (ˆε t ˆε t 1 ) 2 T t=2 ˆε2 t 2(1 ˆρ) ˆρ t 1 t d U d L AR(1)

r ε t = ρ 1 ε t 1 + ρ 2 ε t 2 + + ρ r ε t r + ν t, ν t N(0, σ 2 ν). H 0 : ρ 1 = ρ 2 = = ρ r = 0, H 1 : ρ 1 0 ρ 2 0 ρ r 0. ˆε t R 2 ˆε t = γ 1 + K γ i x it + i=2 r ρ j ˆε t j + ν t, ν t N(0, σν). 2 T j=1 (T r)r 2 χ 2 r. (T r) R 2 T r r (T r) r r R 2 K y t = β 1 + β i x it + ε t, ε t = ρε t 1 + ν t. i=2 1 < ρ < 1 ν t E[ν t ε t 1 ] = 0 (ν t ε t 1 ) = σ 2 ν (ν t, ν s ) = 0 t s ε t = ν t + ρν t 1 + ρ 2 ν t 2 + ρ 3 ν t 3 +.

E[ε t ] = 0, (ε t ) = σ 2 ν + ρ 2 σ 2 ν + = σ2 ν 1 ρ 2. ρ < 1 ρ = 0 σε 2 = σ2 ν 1 ρ 2 CO y t = β 1 + K β i x it + ε t i=2 ε t = ρε t 1 + ν t. ˆε t = ρˆε t 1 + ν t. ˆρ yt = y t ˆρy t 1 β1 = (1 ˆρ)β 1 x 2t = (x 2t ˆρx 2(t 1) K yt = β1 + β i x it + ν t y t = β 1 + i=2 K β i x it + ν t i=2 ˆρ AR(1) AR(q) P W

y t = β 1 + K i=2 β ix it + ε t ˆε t x 2t = α 1 + K i=3 α ix it + r t ˆr t ˆα t = ˆr tˆε t 4(T /100) 2/9 ˆv = T t=1 ˆα2 t + 2 [ ] ( g T h=1 1 h g+1 t=h+1 ˆα t ˆα t h ) g x 2 se( ˆβ 2 ) HAC = ( se( ˆβ ) 2 2 ) ˆv. ˆσ ε x 3 x K y

y t W = T [ b 2 1 6 + (b 2 3) 2 ] 24 T b 1 b 2 b 1 = E[ε3 ] (σ 2 ) 3/2, b 2 = E[ε4 ] (σ 2 ) 2. W χ 2 (2) b 1 b 2 ˆε

ŷ 2 t ŷ 3 t ŷ 4 t y y t = β 1 + β 2 x 1t + + β K x Kt + ε t y t = α 1 + α 2 ŷ 2 t + + α p ŷ p t + K β i x it + ν t. y t F T R 2 χ 2 (p 1) R 2 A B RSS r T 2K A RSS ur,a T A K B RSS ur,b T B K F F = i=1 RSS r (RSS ur,a +RSS ur,b ) K RSS ur,a +RSS ur,b. T 2K F F F (K, T 2K) F F

y y RSS RSS RSS 1 T 1 RSS RSS 1 RSS T 1 K T 2 T 2 F (T 2, T 1 K) F F F F F ±2 ±2

y = Xβ + ε E[ε X] = 0 E[εε X] = σ 2 Ω = Σ, Ω n T ˆβ = (X X) 1 X y = β + (X X) 1 X ε F t ˆβ X [ ˆβ X] = E[( ˆβ β)( ˆβ β) X] = E[(X X) 1 X εε X(X X) 1 X] = (X X) 1 X (σ 2 Ω)X(X X) 1 ( ) 1 ( ) ( ) 1 = σ2 1 1 1 n n X X n X ΩX n X X ( ) 1 ( ) ( ) 1 1 1 1 = n X X n Φ n X X. ( ) Φ = σ2 n X 1 ΩX = n X (y Xβ) E X [[b X]] ˆβ ε ε ˆβ X N(β, σ 2 (X X) 1 (X ΩX)(X X) 1 ) [ ˆβ X] ˆβ (X X/n) 1 (σ 2 /n)(x ΩX/n) Q = p (X X/n) p (X ΩX/n) ˆβ β p ˆβ = β.

X Ω ˆβ Ω [ ˆβ] = σ2 n Q 1 p ( 1 n X ΩX ) Q 1. Ω Ω Ω Ω Ω ˆβ σ 2 Ω ˆβ V OLS = 1 n ( ) 1 ( ) ( ) 1 1 1 1 n X X n X [σ 2 Ω]X n X X. σ 2 Ω = E[εε X] (Ω) = n σ 2 Ω = σ 2 I Σ = (σ ij ) i,j = σ 2 Ω = σ 2 (ω ij ) i,j K(K + 1)/2 Q = 1 n X ΣX = 1 n σ ij x i x n j. x i i X Q i = 1,, K i,j=1 x i1 x i2 X = [x 1,, x K ], x i = x i = [x 1i, x 2i,, x Ki ] x j = [x 1j, x 2j,, x Kj ], 1 i, j n. x in x 1 x 2 X =, X = [ x 1, x 2,..., x n ], X ΣX = x n n σ ij x i x j i,j=1 ˆβ β ˆε i ε i X ˆε Q

S 0 = 1 n ˆε 2 i n x i x i i=1 p S 0 = p Q. [ ˆβ] = 1 ( ) ( ) 1 1 1 n ( ) 1 1 n n X X ˆε 2 i n x i x i n X X = n (X X) 1 S 0 (X X) 1. i=1 ˆβ Q = 1 n σ ij x i x j n ˆQ = 1 n i,j=1 n i,j=1 ˆε iˆε j x i x j ˆQ 1/n n 2 ˆQ X ˆQ ˆQ = S 0 + 1 n L n l=1 t=l+1 w lˆε tˆε t l ( x t x t l + x t l x t), w l = 1 l L + 1. L L T 1/4