KURODA S METHOD FOR CONSTRUCTING CONSISTENT INPUT-OUTPUT DATA SETS. Peter J. Wilcoxen. Impact Research Centre, University of Melbourne.

Similar documents
7.0 Equality Contraints: Lagrange Multipliers

Algorithms behind the Correlation Setting Window

2/20/2013. Topics. Power Flow Part 1 Text: Power Transmission. Power Transmission. Power Transmission. Power Transmission

A New Method for Solving Fuzzy Linear. Programming by Solving Linear Programming

Some Different Perspectives on Linear Least Squares

PRACTICAL CONSIDERATIONS IN HUMAN-INDUCED VIBRATION

D. L. Bricker, 2002 Dept of Mechanical & Industrial Engineering The University of Iowa. CPL/XD 12/10/2003 page 1

Non-degenerate Perturbation Theory

The Mathematics of Portfolio Theory

A Conventional Approach for the Solution of the Fifth Order Boundary Value Problems Using Sixth Degree Spline Functions

2006 Jamie Trahan, Autar Kaw, Kevin Martin University of South Florida United States of America

Parallelized methods for solving polynomial equations

Construction of Composite Indices in Presence of Outliers

A Penalty Function Algorithm with Objective Parameters and Constraint Penalty Parameter for Multi-Objective Programming

Some results and conjectures about recurrence relations for certain sequences of binomial sums.

MEASURES OF DISPERSION

ROOT-LOCUS ANALYSIS. Lecture 11: Root Locus Plot. Consider a general feedback control system with a variable gain K. Y ( s ) ( ) K

SEMI-TIED FULL-COVARIANCE MATRICES FOR HMMS

Assignment 5/MATH 247/Winter Due: Friday, February 19 in class (!) (answers will be posted right after class)

Mu Sequences/Series Solutions National Convention 2014

Capacitated Plant Location Problem:

Chapter 9 Jordan Block Matrices

Standard Deviation for PDG Mass Data

Solving the fuzzy shortest path problem on networks by a new algorithm

1 Onto functions and bijections Applications to Counting

The Selection Problem - Variable Size Decrease/Conquer (Practice with algorithm analysis)

MOLECULAR VIBRATIONS

Coherent Potential Approximation

3D Reconstruction from Image Pairs. Reconstruction from Multiple Views. Computing Scene Point from Two Matching Image Points

Polyphase Filters. Section 12.4 Porat

Assignment 7/MATH 247/Winter, 2010 Due: Friday, March 19. Powers of a square matrix

1 Lyapunov Stability Theory

TESTS BASED ON MAXIMUM LIKELIHOOD

Decomposition of Hadamard Matrices

ε. Therefore, the estimate

Lecture 9: Tolerant Testing

The number of observed cases The number of parameters. ith case of the dichotomous dependent variable. the ith case of the jth parameter

18.413: Error Correcting Codes Lab March 2, Lecture 8

CS 2750 Machine Learning. Lecture 7. Linear regression. CS 2750 Machine Learning. Linear regression. is a linear combination of input components x

( ) 2 2. Multi-Layer Refraction Problem Rafael Espericueta, Bakersfield College, November, 2006

Bounds on the expected entropy and KL-divergence of sampled multinomial distributions. Brandon C. Roy

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 17

Review Exam I Complex Analysis. Cauchy s Integral Formula (#0). Let G be a region in C, let Bar (, ) G and let γ be the circle C(a,r), oriented.

Introduction to local (nonparametric) density estimation. methods

GENERALIZATIONS OF CEVA S THEOREM AND APPLICATIONS

An Innovative Algorithmic Approach for Solving Profit Maximization Problems

Stationary states of atoms and molecules

A unified matrix representation for degree reduction of Bézier curves

Laboratory I.10 It All Adds Up

The theoretical background of

ESS Line Fitting

Lecture 07: Poles and Zeros

Physics 114 Exam 2 Fall Name:

A tighter lower bound on the circuit size of the hardest Boolean functions

Chapter 5 Properties of a Random Sample

hp calculators HP 30S Statistics Averages and Standard Deviations Average and Standard Deviation Practice Finding Averages and Standard Deviations

UNIT 2 SOLUTION OF ALGEBRAIC AND TRANSCENDENTAL EQUATIONS

Summary of the lecture in Biostatistics

L5 Polynomial / Spline Curves

Ahmed Elgamal. MDOF Systems & Modal Analysis

Arithmetic Mean and Geometric Mean

Basic Concepts in Numerical Analysis November 6, 2017

Lecture 7. Confidence Intervals and Hypothesis Tests in the Simple CLR Model

Functions of Random Variables

Department of Agricultural Economics. PhD Qualifier Examination. August 2011

Point Estimation: definition of estimators

A Mean Deviation Based Method for Intuitionistic Fuzzy Multiple Attribute Decision Making

Lecture 12 APPROXIMATION OF FIRST ORDER DERIVATIVES

Johns Hopkins University Department of Biostatistics Math Review for Introductory Courses

Comparing Different Estimators of three Parameters for Transmuted Weibull Distribution

On the energy of complement of regular line graphs

The Mathematical Appendix

Johns Hopkins University Department of Biostatistics Math Review for Introductory Courses

Uniform DFT Filter Banks 1/27

means the first term, a2 means the term, etc. Infinite Sequences: follow the same pattern forever.

d dt d d dt dt Also recall that by Taylor series, / 2 (enables use of sin instead of cos-see p.27 of A&F) dsin

Complete Convergence and Some Maximal Inequalities for Weighted Sums of Random Variables

Derivation of 3-Point Block Method Formula for Solving First Order Stiff Ordinary Differential Equations

C-1: Aerodynamics of Airfoils 1 C-2: Aerodynamics of Airfoils 2 C-3: Panel Methods C-4: Thin Airfoil Theory

ANALYSIS ON THE NATURE OF THE BASIC EQUATIONS IN SYNERGETIC INTER-REPRESENTATION NETWORK

Symmetry of the Solution of Semidefinite Program by Using Primal-Dual Interior-Point Method

CIS 800/002 The Algorithmic Foundations of Data Privacy October 13, Lecture 9. Database Update Algorithms: Multiplicative Weights

Chapter 2 - Free Vibration of Multi-Degree-of-Freedom Systems - II

International Journal of Mathematical Archive-3(5), 2012, Available online through ISSN

Log1 Contest Round 2 Theta Complex Numbers. 4 points each. 5 points each

Analytical Study of Fractal Dimension Types in the Context of SPC Technical Paper. Noa Ruschin Rimini, Irad Ben-Gal and Oded Maimon

For combinatorial problems we might need to generate all permutations, combinations, or subsets of a set.

The Geometric Least Squares Fitting Of Ellipses

Q-analogue of a Linear Transformation Preserving Log-concavity

Global Optimization for Solving Linear Non-Quadratic Optimal Control Problems

Mean is only appropriate for interval or ratio scales, not ordinal or nominal.

Lecture 02: Bounding tail distributions of a random variable

Bayes Estimator for Exponential Distribution with Extension of Jeffery Prior Information

Debabrata Dey and Atanu Lahiri

{ }{ ( )} (, ) = ( ) ( ) ( ) Chapter 14 Exercises in Sampling Theory. Exercise 1 (Simple random sampling): Solution:

Numerical Experiments with the Lagrange Multiplier and Conjugate Gradient Methods (ILMCGM)

X X X E[ ] E X E X. is the ()m n where the ( i,)th. j element is the mean of the ( i,)th., then

SUBCLASS OF HARMONIC UNIVALENT FUNCTIONS ASSOCIATED WITH SALAGEAN DERIVATIVE. Sayali S. Joshi

Rademacher Complexity. Examples

Duality Theory for Interval Linear Programming Problems

Transcription:

KURODA S METHOD FOR CONSTRUCTING CONSISTENT INPUT-OUTPUT DATA SETS by Peter J. Wlcoxe Ipact Research Cetre, Uversty of Melboure Aprl 1989 Ths paper descrbes a ethod that ca be used to resolve cossteces betwee put-output data fro varous sources. It belogs to the class of algorths ds- cussed by Bacharach (1970) ad has bee used extesvely by Kuroda (1988) ad Wlcoxe (1988). I put-output aalyss t s ofte ecessary to use cosstet data sets orgatg fro dfferet goveret ageces. For exaple, the table of terdustry trasactos created by oe agecy ay ot be cosstet wth value-added ad fal dead vectors produced else- agree: a table of terdustry trasactos, a vector of coodty outputs, ad a vector of where. I ths case, the vestgator wll be cofroted wth three peces of data whch do ot gross outputs by dustry. The task the becoes adustg the trasactos table to atch the coodty ad dustry output vectors. I the past, ths proble has bee solved by usg the RAS ethod. RAS s a teratve algorth whch scales the rows ad colus of the trasactos table up ad dow repeatedly utl the table s row ad colu sus agree wth the target vectors. It has bee show that RAS wll evetually coverge, but the result wll ot ecessarly be close ay ecooc sese to the orgal trasactos table. The purpose of ths paper s to defe a easure of how far a ew trasactos table s fro the orgal, ad to derve a algorth whch wll costruct a table zg that dstace.

-- Gve a atrx X of tal data, defe r ad c to be the shares of each eleo et the row ad colu sus of the orgal atrx: r = o X, c = X o (1) X X = 1 o = 1 o Let R be a vector of target row totals, ad C a vector of desred colu totals. The followg fucto ca the be used to easure the dstace betwee a revsed atrx X ad the orgal (eboded r ad c ), where w ad v are arbtrary sets of weghts: Q = 1 =1= 1 X r w + 1 R = 1= 1 c CX v () It s ow possble to choose X to ze ths fucto subect to the followg costrats: R = X (3) = 1 =1 C = X (4) The Lagraga for ths proble s: L = 1 =1= 1 X r w + 1 R = 1= 1 =1 =1 =1 c CX =1 v + λ (R X ) + µ (C X ) (5 )

-3- Takg frst order codtos gves: L = X X r R w + R X c C v λ µ =0 (6) C Collect ters X, ad for coveece ake the followg deftos: S = w v + (7) R C 1 G = r w + R c v (8) C Ths allows the frst order codtos to be rewrtte as show X = S (λ + µ + G ) (9) Both S ad G deped oly o tal data ad the weghts w ad v, so t s oly ecessary to detere λ ad µ, to fd the optal X. These ay be detered by applyg the co strats: R = X = S (λ + µ + G ) (10) = 1 =1 = 1 =1 C = X = S (λ + µ + G ) (11 )

-4- R C Defe dagoal atrces S ad S as dcated: R =1 S = S, S = S (1) C =1 I atrx otato, the costrats ca ow be expressed as R S. λ+s. µ=r A (13) C S. λ+s. µ=c B (14) where A ad B are vectors defed as follows: A = S G B = S G (15) = 1, = 1 Wth ore apulato, t s possble to derve explct forulae for λ ad µ. For coputa- toal purposes, however, t s better to arrage the equatos to the followg syste, whch ca be solved easly by ay copetet uercal package: R S S SS µ λ = R A C C B (16) Ared wth the values of λ ad µ, the optal choce of X ca be coputed drectly.

-5- It s worthwhle to exae a few of the possble weghtg schees that ca be used. The ost obvous approach s to weght all errors equally, whch eas that w =v =1 for all ad. O the other had, the followg choce of weghts results a drastc splfcato of the revso forula: w = 1 R, v = 1 C (17) Ths eas that S =1 for all ad. Moreover, the followg s true of G : G = 1 (r R + c C ) (18) whch s sply the average of the values obtaed by applyg the orgal shares to the target row ad colu sus. Furtherore, t ca be show that all of the followg are true: =1 λ = 0, µ = 0, (19) =1 S R C =, S =. (0) Ths eas that the revso forula has a partcularly sple for: X = G + 1 (R G ) + 1 (C G ) (1) = 1 = 1 Thus, the revsed X s ust G (whch has the terpretato above) adusted to correct the

-6- row ad colu sus. It s portat to ote that ths ethod does ot guaratee that all eleets of X wll be oegatve. Kuroda proposes a dfferet weghtg schee, wth w ad v detered by the followg equatos: w = 1, v = 1. () r c Ths choce of weghts causes Q to be a fucto of the percetage chages the coeffcets: Q = 1 =1 =1 X /R 1 + r 1 =1=1 X /C 1 (3) c I ost cases, ths wll esure that all eleets of X are postve, sce akg oe egatve would requre a chage of ore tha oe hudred percet, resultg a large value of Q. It s possble, however, for egatve ubers to arse f the row ad colu targets dffer substatally fro the correspodg totals of the tal array.

-7- REFERENCES Bachrach, Mchael (1970), Bproportoal Matrces ad Iput-Output Chage, Cabrdge: Cabrdge Uversty Press. Kuroda, M. (1988), "A Method of Estato for the Updatg Trasacto Matrx the Iput-Output Relatoshps," Statstcal Data Bak Systes, K. Uo ad S. Shshdo ( eds.), Asterda: North Hollad. Wlcoxe, P.J. (1988), "The Effects of Evroetal Regulato ad Eergy Prces o U.S. Ecooc Perforace," Ph.D. Thess, Harvard Uversty.