Data Envelopment Analysis with Unbalanced Data Timo Kuosmanen (Wageningen University, The Netherlands) INFORMS Annual Meeting, Atlanta 19-22 October 2003.

Slides:



Advertisements
Similar presentations
Benchmarking Sustainable Development: A Synthetic Meta-Index Approach
Advertisements

Efficiency and Productivity Measurement: Data Envelopment Analysis
Efficiency and Productivity Measurement: Bootstrapping DEA Scores
DEA and Stochastic Dominance Efficiency Analysis of Investment Portfolios: Do Evironmentally Responsible Mutual Funds Diversify Efficiently? Timo Kuosmanen.
Using Stochastic Dominance criteria in Data Envelopment Analysis of mutual funds Timo Kuosmanen Wageningen University, The Netherlands EURO / INFORMS joint.
Duality Theory of Non-Convex Technologies Timo Kuosmanen Wageningen University EEA-ESEM 2003, Stockholm August 2003.
Abstract The influence of the treatment of Multiple Destination Trips (MDT) to the Consumer Surplus (CS) estimates obtained by the Travel Cost Method (TCM)
Weak Disposability in Nonparametric Production Analysis with Undesirable Outputs Timo Kuosmanen Wageningen University, The Netherlands 14th EAERE Annual.
Synthetic Meta-Index of Sustainable Development: A DEA Approach Laurens Cherchye (Catholic University of Leuven, Belgium) Timo Kuosmanen (Wageningen University,
Subsidy measurement and classification: developing a common framework Workshop on Environmentally Harmful Subsidies, Paris, 7-8 November 2002 Ronald Steenblik.
INTRA-INDUSTRY TRADE AND THE SCALE EFFECTS OF ECONOMIC INTEGRATION Elisa Riihimäki Statistics Finland, Business Structures September
1 Alternative measures of well-being Joint work by ECO/ELSA/STD.
Divide-and-Conquer CIS 606 Spring 2010.
6.4 Best Approximation; Least Squares
Monday, October 27, 2003 X-Change Technologies—Compliance proposal 1 Compliance Proposal by X-Change Technologies.
1 A Systematic Review of Cross- vs. Within-Company Cost Estimation Studies Barbara Kitchenham Emilia Mendes Guilherme Travassos.
Performance Evaluation and Benchmarking Using DEA
General Linear Model With correlated error terms  =  2 V ≠  2 I.
Diversification in the Stochastic Dominance Efficiency Analysis Timo Kuosmanen University of Copenhagen, Denmark Wageningen University, The Netherlands.
Copyright © 2006 Pearson Education Canada Inc Course Arrangement !!! Nov. 22,Tuesday Last Class Nov. 23,WednesdayQuiz 5 Nov. 25, FridayTutorial 5.
1 Software Testing and Quality Assurance Lecture 31 – Testing Systems.
Exam 1 – 115a. Basic Probability For any event E, The union of two sets A and B, A  B, includes items that are in either A or B. The intersection, A.
Sequence Alignment II CIS 667 Spring Optimal Alignments So we know how to compute the similarity between two sequences  How do we construct an.
Rank Robustness of Composite Indices: Dominance and Ambiguity James E. Foster George Washington University & Oxford Mark McGillivray Ausaid Suman Seth.
Arrays. The array data structure An array is an indexed sequence of components –Typically, the array occupies sequential storage locations –The length.
Lecture 10 Comparison and Evaluation of Alternative System Designs.
Incomplete Block Designs
1 1 Slide © 2003 South-Western/Thomson Learning™ Slides Prepared by JOHN S. LOUCKS St. Edward’s University.
Data Envelopment Analysis (DEA). Which Unit is most productive? DMU = decision making unit DMU labor hrs. #cust
Socially Responsible Investing (SRI) Value-Based or “Ethical” mutual funds: –Create screens to prevent investment in organizations that promote or participate.
Annex I: Methods & Tools prepared by some members of the ICH Q9 EWG for example only; not an official policy/guidance July 2006, slide 1 ICH Q9 QUALITY.
Measuring Electricity Generation Efficiency Data Envelopment Analysis versus Fixed Proportion Technology Indicators.
Objectives -To understand the steps in generating the MATRIX and MONITOR. -To show that the MATRIX and MONITOR can reveal the types of health equity problems.
Retail Labor Planning Model – Alix Partners Carolyn Taricco Erin Gripp Victoria Cohen.
An evaluation of European airlines’ operational performance.
And, now take you into a WORLD of……………...
A two-stage approach for multi- objective decision making with applications to system reliability optimization Zhaojun Li, Haitao Liao, David W. Coit Reliability.
Bo Sjo. Development per Aid Dollar Data Envelopment Analysis applied to aid efficiency Bo Sjö March 2009.
1 Helsinki University of Technology Systems Analysis Laboratory INFORMS 2007 Seattle Efficiency and Sensitivity Analyses in the Evaluation of University.
Compuware Corporation Developing the most appropriate Verification Strategy A presentation for CMMI Technology Conference Denver, Colorado November 20,
1 DEA Based Approaches and Their Applications in Supply Chain Management Dr. Sri Talluri Professor of Supply Chain Management Presentation at the Helsinki.
Statistical Inference Statistical Inference involves estimating a population parameter (mean) from a sample that is taken from the population. Inference.
BUSI 6480 Lecture 8 Repeated Measures.
MODEL FOR DEALING WITH DUAL-ROLE FACTORS IN DEA: EXTENSIONS GONGBING BI,JINGJING DING,LIANG LIANG,JIE WU Presenter : Gongbing Bi School of Management University.
Spectral Sequencing Based on Graph Distance Rong Liu, Hao Zhang, Oliver van Kaick {lrong, haoz, cs.sfu.ca {lrong, haoz, cs.sfu.ca.
Constructing the Welfare Aggregate Part 2: Adjusting for Differences Across Individuals Salman Zaidi Washington DC, January 19th,
Unit 1 MATRICES Dr. Shildneck Fall, WHAT IS A MATRIX? A Matrix is a rectangular array of numbers placed inside brackets. A Matrix is a rectangular.
Data Envelopment Analysis
QUANTITATIVE TECHNIQUES
MAVILLE ALASTRE-DIZON Philippine Normal University
Schedule Reading material for DEA: F:\COURSES\UGRADS\INDR\INDR471\SHARE\reading material Homework 1 is due to tomorrow 17:00 ( ). Homework 2 will.
Table of Contents Matrices - Definition and Notation A matrix is a rectangular array of numbers. Consider the following matrix: Matrix B has 3 rows and.
ESTIMATING WEIGHT Course: Special Topics in Remote Sensing & GIS Mirza Muhammad Waqar Contact: EXT:2257 RG712.
Benchmarking for Improved Water Utility Performance.
Copyright 2010, The World Bank Group. All Rights Reserved. Producer prices, part 2 Measurement issues Business Statistics and Registers 1.
Comparison of Estimation Methods for Agricultural Productivity Yu Sheng ABARES the Superlative vs. the Quantity- based Index Approach August 2015.
Labour Cost Index (LCI) Calculation of the LCI in Denmark.
Leanness Score of Value Stream Maps 1. ANES WEDARING DHARMA ( ) 2. EKA NOVANTARA ( ) 3. RENAN LINTANG PRAKOSO ( )
Indexes, Scales, and Typologies
Benchmarking of Indian Urban Water Sector: Performance Indicator System versus Data Envelopment Analysis By: Dr. Mamata Singh, Dr. Atul K. Mittal, and.
Unit 1: Matrices Day 1 Aug. 7th, 2012.
General principles in building a predictive model
§2-3 Observability of Linear Dynamical Equations
Life cycle patterns, farm performance and structural change: an empirical research Steven Van Passel I’m working for the policy research centre for sustainable.
SUTs – data sources and bridge tables
Arrays.
Fourth Lisbon Research Workshop on Economics, Statistics and Econometrics of Education January 26th to 26th 2017 Performance indicators and rankings in.
Statistics and Data Analysis
Chapter 3 Central tendency and variation
P 72 (PDF 76) Figure 32 Information item name Rules in columns
Presentation transcript:

Data Envelopment Analysis with Unbalanced Data Timo Kuosmanen (Wageningen University, The Netherlands) INFORMS Annual Meeting, Atlanta October 2003

Unbalanced data? Suppose output j of DMU k is missing (unavailable).

Unbalanced data? Suppose output j of DMU k is missing (unavailable). Usual approach is to restore a balanced output matrix by excluding DMU k

Unbalanced data? Suppose output j of DMU k is missing (unavailable). Usual approach is to restore a balanced output matrix by excluding DMU k

Unbalanced data? Suppose output j of DMU k is missing (unavailable). Usual approach is to restore a balanced output matrix by excluding DMU k excluding output j

Problems Both approaches involve a loss of information about production possibilities in observed outputs of discarded DMU k observed values of excluded output j The choice to exclude either DMU or output influences the results Criteria for excluding rows/columns are typically not explicitly reported

Proposition Why dont we simply tolerate the missing piece of data and denote the missing output value by zero (0)? Zero is the theoretical lower bound for output values. No technical reason for including 0 outputs in DEA.

Notation Define the following production possibility sets: T DMU : exclude the DMU with missing value T Y : exclude the output with missing value T UB : denote missing output by 0 T IDEAL : ideal case where all data are available

Main Theorem Production possibility sets T UB, T IDEAL, T DMU, and T Y are nested in the sense that

Example (2 outputs, 5 DMUs)

Influence on efficiency scores Theorem 2: For DMU k with missing value of output j, using unbalanced data and eliminating output j yield equal DEA efficiency scores. Theorem 3: For DMU l with complete data, using unbalanced data can only yield worse efficiency score than excluding DMU k with missing data from the reference set.

Equity issues The unbalanced DEA model imposes DMUs with missing outputs more stringent efficiency criteria might be viewed unfair incentives for collecting & reporting data Even if we exclude DMUs with missing outputs from efficiency comparisons / rankings, there is no harm in including them in the reference technology! Might adjust the efficiency scores to take into account differences in dimensionality across DMUs?

Extensions Missing inputs can be handled analogously by labeling blank entries by some big M. Weight restrictions can interfere with the results in unintended way. We may relax weight restrictions by writing them as >

Case study: Sustainable Development indices Cherchye & Kuosmanen (2002) use DEA to construct a meta-index of Sustainable Development (SD) from 14 (SD) indicators for 154 countries. The 14x143 data matrix contains 2156 elements, of which 18% (= 395 elements) were missing. Complete data available only for 14 countries.

Comparison of approaches

Conclusions A first systematic attempt to analyze the effects of eliminating missing values Keeping blank entries in the output data can only improve estimation of the production frontier. Differences in dimensionality across DMUs can be unfair for DMUs with good performance in missing outputs Research question: Can a fair handicap system be constructed for making efficiency scores better comparable if dimensionality differs across DMUs???

Want to read more? Full paper can be downloaded from my homepage: Or send to: