Spatial Regression. 10. Specification Tests (2) Luc Anselin. Copyright 2017 by Luc Anselin, All Rights Reserved

Similar documents
Lecture 6: Hypothesis Testing

Spatial Regression. 9. Specification Tests (1) Luc Anselin. Copyright 2017 by Luc Anselin, All Rights Reserved

Spatial Regression. 3. Review - OLS and 2SLS. Luc Anselin. Copyright 2017 by Luc Anselin, All Rights Reserved

Econometrics Part Three

SPATIAL ECONOMETRICS: METHODS AND MODELS

Econometrics. Week 8. Fall Institute of Economic Studies Faculty of Social Sciences Charles University in Prague

Tests forspatial Lag Dependence Based onmethodof Moments Estimation

Spatial Regression. 14. Spatial Panels (2) Luc Anselin. Copyright 2017 by Luc Anselin, All Rights Reserved

LECTURE 11. Introduction to Econometrics. Autocorrelation

Spatial Regression. 15. Spatial Panels (3) Luc Anselin. Copyright 2017 by Luc Anselin, All Rights Reserved

Testing Random Effects in Two-Way Spatial Panel Data Models

ON THE NEGATION OF THE UNIFORMITY OF SPACE RESEARCH ANNOUNCEMENT

Bootstrap Test Statistics for Spatial Econometric Models

Luc Anselin and Nancy Lozano-Gracia

Spatial Analysis 2. Spatial Autocorrelation

School of Mathematical Sciences. Question 1

Working Paper No Introduction to Spatial Econometric Modelling. William Mitchell 1. April 2013

Spatial Autocorrelation and Interactions between Surface Temperature Trends and Socioeconomic Changes

Eksamen på Økonomistudiet 2006-II Econometrics 2 June 9, 2006

Outline. Overview of Issues. Spatial Regression. Luc Anselin

Repeated observations on the same cross-section of individual units. Important advantages relative to pure cross-section data

LECTURE 10: MORE ON RANDOM PROCESSES

Heteroskedasticity. Part VII. Heteroskedasticity

Økonomisk Kandidateksamen 2004 (I) Econometrics 2. Rettevejledning

Multiple Regression Analysis: Heteroskedasticity

RAO s SCORE TEST IN SPATIAL ECONOMETRICS

1 Motivation for Instrumental Variable (IV) Regression

Outline ESDA. Exploratory Spatial Data Analysis ESDA. Luc Anselin

1. You have data on years of work experience, EXPER, its square, EXPER2, years of education, EDUC, and the log of hourly wages, LWAGE

Testing for Spatial Group Wise Testing for SGWH. Chasco, Le Gallo, López and Mur, Heteroskedasticity.

IV Estimation and its Limitations: Weak Instruments and Weakly Endogeneous Regressors

SPACE Workshop NSF NCGIA CSISS UCGIS SDSU. Aldstadt, Getis, Jankowski, Rey, Weeks SDSU F. Goodchild, M. Goodchild, Janelle, Rebich UCSB

MULTIPLE REGRESSION AND ISSUES IN REGRESSION ANALYSIS

Review on Spatial Data

GeoDa and Spatial Regression Modeling

Least Absolute Value vs. Least Squares Estimation and Inference Procedures in Regression Models with Asymmetric Error Distributions

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

Regional Science and Urban Economics

Econ 510 B. Brown Spring 2014 Final Exam Answers

Reading Assignment. Serial Correlation and Heteroskedasticity. Chapters 12 and 11. Kennedy: Chapter 8. AREC-ECON 535 Lec F1 1

Testing Panel Data Regression Models with Spatial Error Correlation*

Spatial Regression. 1. Introduction and Review. Luc Anselin. Copyright 2017 by Luc Anselin, All Rights Reserved

Economics 308: Econometrics Professor Moody

A Practitioner s Guide to Cluster-Robust Inference

Introduction Large Sample Testing Composite Hypotheses. Hypothesis Testing. Daniel Schmierer Econ 312. March 30, 2007

MS&E 226: Small Data

Economics 582 Random Effects Estimation

Christopher Dougherty London School of Economics and Political Science

Dynamic Panels. Chapter Introduction Autoregressive Model

Economics 536 Lecture 7. Introduction to Specification Testing in Dynamic Econometric Models

G. S. Maddala Kajal Lahiri. WILEY A John Wiley and Sons, Ltd., Publication

ECON2228 Notes 10. Christopher F Baum. Boston College Economics. cfb (BC Econ) ECON2228 Notes / 48

Linear Regression with Time Series Data

Reasons for Instability in Spatial Dependence Models

Chapter 4: Constrained estimators and tests in the multiple linear regression model (Part III)

Spatial Regression. 11. Spatial Two Stage Least Squares. Luc Anselin. Copyright 2017 by Luc Anselin, All Rights Reserved

LECTURE 10: NEYMAN-PEARSON LEMMA AND ASYMPTOTIC TESTING. The last equality is provided so this can look like a more familiar parametric test.

Proceedings of the 8th WSEAS International Conference on APPLIED MATHEMATICS, Tenerife, Spain, December 16-18, 2005 (pp )

matrix-free Hypothesis Testing in the Regression Model Introduction Kurt Schmidheiny Unversität Basel Short Guides to Microeconometrics Fall 2018

Introductory Econometrics

Size and Power of the RESET Test as Applied to Systems of Equations: A Bootstrap Approach

ECON2228 Notes 10. Christopher F Baum. Boston College Economics. cfb (BC Econ) ECON2228 Notes / 54

Econometrics of Panel Data

y it = α i + β 0 ix it + ε it (0.1) The panel data estimators for the linear model are all standard, either the application of OLS or GLS.

ECON 312 FINAL PROJECT

Spatial Effects in Convergence of Portuguese Product

WISE MA/PhD Programs Econometrics Instructor: Brett Graham Spring Semester, Academic Year Exam Version: A

Analyzing spatial autoregressive models using Stata

Econometrics of Panel Data

A SPATIAL CLIFF-ORD-TYPE MODEL WITH HETEROSKEDASTIC INNOVATIONS: SMALL AND LARGE SAMPLE RESULTS

(a) (3 points) Construct a 95% confidence interval for β 2 in Equation 1.

Econometrics II - EXAM Answer each question in separate sheets in three hours

A discussion on multiple regression models

Quick Review on Linear Multiple Regression

F9 F10: Autocorrelation

A Robust LM Test for Spatial Error Components

Linear Regression with Time Series Data

Final Exam. Question 1 (20 points) 2 (25 points) 3 (30 points) 4 (25 points) 5 (10 points) 6 (40 points) Total (150 points) Bonus question (10)

Spatial Effects in Convergence of Portuguese Product

Spatial Regression. 6. Specification Spatial Heterogeneity. Luc Anselin.

Warwick Economics Summer School Topics in Microeconometrics Instrumental Variables Estimation

Non-Stationary Time Series and Unit Root Testing

y ˆ i = ˆ " T u i ( i th fitted value or i th fit)

Iris Wang.

Reliability of inference (1 of 2 lectures)

Stat 231 Exam 2 Fall 2013

Functional Form. Econometrics. ADEi.

Intermediate Econometrics

Likely causes: The Problem. E u t 0. E u s u p 0

ECON 4551 Econometrics II Memorial University of Newfoundland. Panel Data Models. Adapted from Vera Tabakova s notes

Specification testing in panel data models estimated by fixed effects with instrumental variables

Non-Stationary Time Series and Unit Root Testing

Binary Dependent Variables

Estimation and Hypothesis Testing in LAV Regression with Autocorrelated Errors: Is Correction for Autocorrelation Helpful?

Economic modelling and forecasting

Wooldridge, Introductory Econometrics, 2d ed. Chapter 8: Heteroskedasticity In laying out the standard regression model, we made the assumption of

splm: econometric analysis of spatial panel data

Introduction to Eco n o m et rics

Econometrics Summary Algebraic and Statistical Preliminaries

Lecture 1: intro. to regresions

Transcription:

Spatial Regression 10. Specification Tests (2) Luc Anselin http://spatial.uchicago.edu 1

robust LM tests higher order tests 2SLS residuals specification search 2

Robust LM Tests 3

Recap and Notation LM-Error test LM-Lag test with with 4

What are the Robust Tests? Problem: both LM-Error and LM-Lag have power against the other alternative LM-Error rejects null in presence of lag model LM-Lag rejects null in presence of error model robust forms of the test make an asymptotic adjustment to correct for this (Anselin et al 1996) 5

Robust LM Tests Robust LM-Error Robust LM-Lag 6

Use of Robust Tests only use when BOTH LM-Error and LM-Lag reject the null do NOT use when neither LM-Error nor LM-Lag are significant select model with most significant statistic = specification search 7

Robust LM-Lag = 3.0 Robust LM-Error = 5.1 only Robust LM-Error rejects the null hypothesis 8

Higher Order Tests 9

Test on SARMA test on higher order alternative, BOTH lag and error dependence H0: λ = ρ = 0 Test χ2 (2) two degrees of freedom (one for lag, one for error) 10

Non-Standard Result test is not the sum of two one-directional tests Test = LM-Error + Robust LM-Lag = LM-Lag + Robust LM-Error 11

> > < < 56.43 = 51.36 + 5.07 56.43 = 53.45 + 2.98 12

Interpretation of SARMA Test caution! SARMA test will be significant when either LM- Error or LM-Lag is highly significant does NOT mean the alternative is a higher order model higher order only makes sense as alternative when evidence of other form of misspecification remains in lower order model (e.g., remaining error autocorrelation in a spatial lag model) 13

2SLS Residuals 14

Generalized Moran s I Test extend to residuals other than OLS residuals from a 2SLS regression no longer a maximum likelihood framework requires explicit CLT 15

Moran s I for 2SLS Residuals Anselin and Kelejian (1997) apply Moran s I requires generalized expression for variance φ2 16

Moran s I for 2SLS Residuals (2) expression for I is standard, using 2SLS residuals expression for variance is complex 17

Special Case - No Spatial Lag standard endogenous variable case test statistic simplifies to LM-Error like expression using 2SLS residuals 18

Specification Search 19

Principle 20

Two Strategies (Florax, Folmer, Rey 2003) forward step-wise strategy move from simple to complex model backward step-wise strategy move from complex to simple model (Hendry) 21

Forward Step-Wise Strategy start from constrained model = non-spatial model (OLS estimation) use LM test statistics to guide model selection problem: pre-testing, due to multiple tests the p-values become suspect corrections for pre-testing are complex 22

Backward Step-Wise Strategy start from unconstrained model and test constraints (Hendry approach) proceed from complex to simpler specification problem: requires estimation of complex spatial models first to test parameter constraints (using Wald or LR tests) 23

Specification Search - Step 1 use LM tests first (NOT the robust version) none significant: proceed with OLS LM-Error only significant: spatial error model LM-Lag only significant: spatial lag model both significant: proceed to robust LM tests 24

Specification Search - Step 2 Robust LM-Error significant, Robust LM-Lag is not: spatial error model Robust LM-Lag significant, Robust LM-Error is not: spatial lag model both Robust LM tests are significant: alternative is the most significant (largest value) possibility of higher order model or alternative specifications 25

26

Examples 27

Case 1: No Evidence of Spatial Autocorrelation Columbus crime example with center-periphery indicator variable (= spatial heterogeneity) columbus.shp with columbus_rk.gal CRIME on INC, HOVAL, CP 28

regression results 29

< none of the LM statistics are significant but Moran s I is somewhat, why? < possibly power against heteroskedasticity 30

Case 2: Clear Indication of One Type of Spatial Autocorrelation - Spatial Error Columbus crime example with DISCBD as one of the explanatory variables columbus.shp with columbus_d.gwt (distancebased weights) CRIME on INC, HOVAL, DISCBD 31

regression results 32

<<< < Moran s I significant at p = 0.01 LM-Error significant at p = 0.06 Conclusion: Spatial Error Model 33

why is Moran s I so significant? non-normality? heteroskedasticity? 34

Case 3: Clear Indication of One Type of Spatial Autocorrelation - Spatial Lag Columbus crime example with PLUMB as one of the explanatory variables columbus.shp with columbus_d.gwt (distancebased weights) CRIME on INC, HOVAL, PLUMB 35

regression results 36

> <<< > LM-Lag significant at p=0.01 LM-Error not significant Conclusion: Spatial Lag Model 37

Case 4: LM Tests Not Significant, but one of the Robust LM Tests is Significant columbus.shp with columbus_rk.gal (rook contiguity) CRIME on INC, HOVAL, NSB, DISCBD 38

regression results 39

< < Moran s I weakly significant LM Tests NOT significant Robust LM Error somewhat significant = IGNORE 40

why is Moran s I weakly significant non-normality? heteroskedasticity? 41

Case 5: Both LM-Error and LM-Lag significant, Robust LM-Error Significant south.shp with south_rk.gal (rook contiguity) HR90 on RD90, PS90, UE90, DV90, MA90 42

regression results 43

>>> >>> < <<< 44

Case 6: Both LM-Error and LM-Lag significant, Robust LM-Lag Significant south.shp with south_q.gal (rook contiguity) HR70 on RD70, PS70, UE70, DV70, MA70 45

regression results 46

>>> >>> <<< < 47