LASSO variable selection in data envelopment analysis with small datasets

The curse of dimensionality problem arises when a limited number of observations are used to estimate a high-dimensional frontier, in particular, by data envelopment analysis (DEA). The study conducts a data generating process (DGP) to argue the typical “rule of thumb” used in DEA, e.g. the required number of observations should be at least larger than twice of the number of inputs and outputs, is ambiguous and will produce large deviations in estimating the technical efficiency. To address this issue, we propose a Least Absolute Shrinkage and Selection Operator (LASSO) variable selection technique, which is usually used in data science for extracting significant factors, and combine it in a sign-constrained convex nonparametric least squares (SCNLS), which can be regarded as DEA estimator. Simulation results demonstrate that the proposed LASSO-SCNLS method and its variants provide useful guidelines for the DEA with small datasets.     

拟凹生产函数的分区域估计

本文提出了规模报酬递增生产前沿面的概念并证明了以下两个结论:(1)基于样本数据的DEA生产投入集可划分为规模报酬递增、不变和递减区域;(2)C-D生产函数是拟凹函数,且在规模报酬递增区域非凹,在规模递减区域严格凹。基于上述结论及对生产函数曲面,BCC生产前沿面,规模报酬递增生产前沿面的相互关系的分析,提出了一种生产函数分区域估计方法:在对样本数据(决策单元)依据规模报酬增减性进行分类的基础上进行投入可能集的分划,进而在规模报酬递减(不变)和递增区域上分别通过决策单元的BCC弱有效投影和规模报酬递增弱有效投影估计生产函数。文末,通过实例验证了估计方法的有效性。     

A modified DEA model to estimate the importance of objectives with an application to agricultural economics

This paper demonstrates a connection between data envelopment analysis (DEA) and a non-interactive elicitation method to estimate the weights of objectives for decision-makers in a multiple attribute approach. This connection gives rise to a modified DEA model that allows us to estimate not only efficiency measures but also preference weights by radially projecting each unit onto a linear combination of the elements of the payoff matrix (which is obtained by standard multicriteria methods). For users of multiple attribute decision analysis the basic contribution of this paper is a new interpretation in terms of efficiency of the non-interactive methodology employed to estimate weights in a multicriteria approach. We also propose a modified procedure to calculate an efficient payoff matrix and a procedure to estimate weights through a radial projection rather than a distance minimization. For DEA users, we provide a modified DEA procedure to calculate preference weights and efficiency measures that does not depend on any observations in the dataset. This methodology has been applied to an agricultural case study in Spain.     

Applications of sequential set partitioning: a set of technical information systems problems

Many problems that appear in different contexts are conceptually similar and so are amenable to solution by a common technique. Three such technical Information Systems (IS) problems are: (1) segmentation of computer programs; (2) attribute discretization for decision tree induction; and (3) design of hash tables in database systems. In this paper we show how each of these seemingly different problems can be formulated as a sequential (set) partitioning problem, and solved using a parametric linear programming (LP) procedure. This approach provides optimal solutions unlike previous solution approaches which were either greedy heuristics or limited to solving only a specific problem situation. Given the likelihood that other applications of the sequential partitioning problem exist in IS, the material presented here could be useful to other researchers in formulating the problem at an appropriate level of abstraction so that available optimal solution approaches can be identified. In addition to providing a common solution method, parametric LP frees the user from having to make premature decisions regarding the number of groups for the partition, and this decision can be delayed post solution.     

Measuring efficiency with products,by-products and parent-offspring relations: A conditional two-stage DEA model

In the usual data envelopment analysis (DEA) setting, as pioneered by Charnes et al. (1978) [1], it is assumed that a set of decision making units (DMUs) is to be evaluated in terms of their relative efficiencies in converting a bundle of inputs into a bundle of outputs. The usual assumption in DEA is that each output is impacted by each and every member of the input set. One particular area of recent research is that relating to partial input to output impacts where the main issue addressed is that in many settings not all inputs impact all outputs. In that situation the authors view the DMU as consisting of a set of mutually exclusive subunits, with each subunit having its own unique bundle of inputs and outputs. Examined as well in this area, is the presence of multiple processes for generating sets of outputs. Missing from that earlier work is consideration of the presence of outputs in the form of by-products, giving rise to a parent-offspring phenomenon. One of the modeling complications there is that the parent assumes two different roles; as an input affecting the offspring, while at the same time being the dominant output. This gives rise to a model that we refer to as conditional two-stage. Another complication is that in the presence of multiple processes, by-products often arise out of only a subset of those processes. In the current paper we develop a DEA-type of methodology to handle partial input to output impacts in the presence of by-products.     

存在保证域的模糊非径向偏好DEA模型——基于中科院24个研究所的实证分析

数据包络分析（DEA）是一种非参数化的方法,用于评价具有类似输入和输出的决策单元的效率。传统的非径向DEA模型假设输入和输出数据均为准确值,且对权重变量不加以限制,本文构建了存在保证域的模糊非径向偏好DEA模型,并给出了一种基于模糊数截集的模型求解方法,有效地解决了输入和输出全部或部分为模糊数的决策单元评价问题。最后给出了一个中科院研究所效率评价的实例说明了方法的有效性。     

A fair evaluation of certain stage in a two-stage structure: revisiting the typical two-stage DEA approaches

The efficiency of decision processes with a two-stage structure has been studied by some modified versions of Data Envelopment Analysis (DEA) methodology, such as the relational or centralized model, and the non-cooperative model. After revisiting the rationale of the centralized model provided by the literature, we find that some unfairness exist in its efficiency evaluation of certain stage. This unfairness leads to the usual underestimation of the overall efficiency by the centralized model. Furthermore, because the independent DEA model for one stage ignores its relation and coordination with the other stage and the two-stage system, externalities between these members may arise and lead to the seeming contradiction in efficiencies derived by independent DEA models for the stages and the black box system. Therefore, we argue that the efficiency of certain stage in the context of a two-stage structure should be reevaluated instead of simply using the independent DEA model. A sequence of leader-follower procedures, with data adjustment on intermediates, is proposed to eliminate the externalities and to ensure a fair evaluation. We find that, after this data adjustment, the reevaluation of the second stage for a given Decision Making Unit (DMU) yields the same result as obtained by the standard DEA model regarding the two-stage as a black box. Moreover, some explicit relations are established between the black box model, the centralized model, and the non-cooperative model. Two typical examples taken from the literature illustrate our main results. Our findings also imply that more emphasis should be placed on the game theoretic DEA approach to model the efficiency evaluation of two-stage processes.     

非合作博弈两阶段生产系统的环境效率评价

本文研究了环境效率评价中广泛存在的一类典型问题:两阶段非合作博弈的效率评价问题,该问题有两个明显的特点:(1)决策单元由经济生产子系统和污染物处理子系统两个阶段的子系统构成,(2)经济生产子系统产生的污染物由污染物处理子系统进行处理,且经济生产子系统居于主导地位,而污染物处理子系统居于从属地位.文章在分析各个决策单元整体效率与其子系统效率之间关系的基础上,提出一种能够同时评价决策单元整体效率和其子系统效率的DEA方法;然后引入非合作博弈思想,提出了基于两阶段非合作博弈的环境效率评价的DEA方法;最后采用中国各地区的工业系统的环境效率评价的实例分析说明了该方法的合理性和有效性.由于文章提出的方法能够有效地分析决策单元内部子系统的有效性水平,能够反映子系统之间的主从关系,能够发掘影响决策单元效率的关键因素,因而能够给出更为有效的环境管理信息,为决策者的决策提供科学的依据.     

Joint Models for Toxicology Studies with Dose-Dependent Number of Implantations

Many chemicals interfere with the natural reproductive processes in mammals. The chemicals may prevent the fertilization of an egg or keep a zygote from implanting in the uterine wall. For this reason, toxicology studies with pre-implantation exposure often exhibit a dose-related trend in the number of observed implantations per litter. Standard methods for analyzing developmental toxicology studies are conditioned on the number of implantations in the litter and therefore cannot estimate this effect of the chemical on the reproductive process. This article presents a joint modeling approach to estimating risk in toxicology studies with pre-implantation exposure. In the joint modeling approach, both the number of implanted fetuses and the outcome of each implanted fetus is modeled. Using this approach we show how to estimate the overall risk of a chemical that incorporates the risk of lost implantation due to pre-implantation exposure. Our approach has several distinct advantages over previous methods: (1) it is based on fitting a model for the observed data and, therefore, diagnostics of model fit and selection apply; (2) all assumptions are explicitly stated; and (3) it can be fit using standard software packages We illustrate our approach by analyzing a dominant lethal assay data set (Luning et al., 1966, Mutation Research, 3, 444-451) and compare ourresults with those of Rai and Van Ryzin (1985, Biometrics, 41,1-9) and Dunson (1998, Biometrics, 54, 558-569). In a simulation study, our approach has smaller bias and variance than the multiple imputation procedure of Dunson.     

非合作博弈两阶段生产系统DEA并购效率评价

如何选择并购对象及预测并购可行性是企业管理者在并购决策过程中常面临的问题。针对企业并购预测问题,本文以两阶段生产系统的决策单元为研究对象,研究数据包络分析方法在预测并购可行性方面的应用。该生产系统有两个明显的特点:(1)决策单元由两个阶段串行子系统构成,(2)两个子系统一个处于主导地位,而另一个处于从属地位。两个或者两个以上决策单元并购为一个虚拟决策单元。本文在分析由决策单元整体效率与其子系统效率之间关系的基础上,引入非合作博弈的思想,提出了基于非合作博弈的DEA模型评估虚拟决策单元两阶段生产系统的并购效率:即在保持现有的产出水平和生产效率的前提下,分别求解虚拟决策单元及其子系统的并购效率,分析虚拟决策单元如何通过两个子系统实现整个系统的成本节约。最后,将模型应用于台湾24家非寿险保险公司的仿真分析。由于文章提出的方法能够有效地分析虚拟决策单元内部子系统的并购有效性水平,同时考虑子系统之间的主从关系,因此,能够发掘影响虚拟决策单元并购效率的内部因素,因而能够为管理者并购决策提供更精确有效的管理信息,提高并购决策的科学性。     

SOME DIFFICULTIES AND IMPROVEMENTS IN APPLYING LINEAR PROGRAMMING FORMULATIONS TO THE DISCRIMINANT PROBLEM

In this paper, we discuss some disturbing features of two linear programming (LP) approaches to the discriminant problem. Specifically, we show that both approaches are sensitive to the choice of origin for the data although, intuitively, placement of origin should have no effect on the method of assigning cases to groups. In addition, we show that these LP approaches may lead to discriminant functions which assign all cases to the same group. We show that the usual statistical approach to this problem does not share these difficulties, and we make recommendations for implementing these LP approaches which help to alleviate the difficulties.     

Benchmarking firm performance from a multiple-stakeholder perspective with an application to Chinese banking

We know very little about how a firm would compare against its peers when evaluated from a multiple-stakeholder perspective where the same variables are interpreted differently. Since most medium-to-large organizations acknowledge the multi-dimensional nature of their operations, finding out the performance evaluations of various stakeholders can inform managerial decision-making. Thus, the main motivation for this study is to capture the interactions among different perceptions on a common set of performance measures. Using data envelopment analysis (DEA), we work with an approach that allows a flexible designation of inputs and outputs based on varying perspectives of five key stakeholders in banking. The versatile approach demonstrates that different views from the stakeholder universe can be summarily captured in DEA scores. A numerical example on Chinese commercial banks identifies the compliant (efficient) banks versus rigid (inefficient) banks, as well as the amenable stakeholders (those evaluating banks as efficient) versus the recalcitrant stakeholders (those evaluating banks as inefficient). The aligned views held by management and shareholders as evidenced by significant correlation among performance rankings imply reduced agency costs. Similarly, shared perceptions between customers and employees may encourage management to examine how this important business interface can be improved.     

Ranking using η-efficiency and relative size measures based on DEA

In this paper, we propose a new way to define the size of organizations or Decision Making Units (DMUs) based on Data Envelopment Analysis (DEA). A great number of managerial decisions, social and economic activities, political analysis and engineering process include in their analysis some information about the size of corporations. Usually, organizations consider particular attributes such as revenues, employees, number of clients, profit and so forth to evaluate their sizes, which could result on an incomplete and/or partial size evaluation. Here, we investigate this question more deeply and propose a non-parametric approach for ranking size. To address this issue, we first construct a DEA super-efficiency target vector defined over an extended Technology set and find a common set of weights to determine new efficiencies and sizes, possibly greater than one. After, we define rotation and translation of DMUs and show how to construct a range area, associated to η-efficiencies and relative sizes that are less than or equal to one. Additionally, based on managerial decisions, we show how to use these ideas to achieve new η-efficiency and relative size goals for one or more DMUs in the set. We consider an empirical application to illustrate our approach on a dataset of international airlines and conclude the paper with some final remarks and future directions.     

Eco-efficiency analysis of paper mills along the Huai River: An extended DEA approach

The objective of this paper is to estimate the ecological efficiency of paper mills along the Huai River in China. The main characteristic of the ecological efficiency evaluation problem is that an undesirable output of biochemical oxygen demand (BOD) and a non-discretionary input (BOD emission quota) should be considered simultaneously. By analyzing the impacts of the non-discretionary input on decision-making units’ (DMUs) desirable and undesirable outputs, a non-radial output-oriented DEA model is proposed. In the proposed model, we describe a new approach of defining reference set that requires reference units operate in a similar environment on average. We employ the model to provide efficient inputs/outputs targets for DMU managers to improve DMUs’ efficiencies. Based on the developed model, impacts of the non-discretionary input on DMUs’ returns are also analyzed. We illustrate the proposed model, using real data, for 32 paper mills along the Huai River in China.     

Two-stage DEA models with undesirable input-intermediate-outputs

Data envelopment analysis (DEA) is a non-parametric approach for measuring the relative efficiencies of peer decision making units (DMUs). Many studies have examined DEA efficiencies of two-stage systems, where all the outputs from the first stage are the only inputs to the second stage. Although single-stage DEA models with undesirable input-outputs have been extensively studied, there still lacks of more systematical investigation on two-stage DEA with undesirable variables. For instance, depending on its operating model, even whether an intermediate variable is desirable or undesirable can be questionable for a particular two-stage system. Furthermore, most of the existing studies on two-stage systems focus on the case where only the final outputs are undesirable. In this work, we try to systematically examine two-stage DEA models with undesirable input-intermediate-outputs. Particularly, we utilize the free-disposal axioms to construct the production possibility sets (PPS) and the corresponding DEA models with undesirable variables. The proposed models are then used to illustrate some theoretical perspectives by using the data of China׳s listed banks.     

基于偏好DEA模型的中国纺织业效率评价

基于具有输入输出指标偏好信息的数据包络分析模型评价中国纺织工业的效率,给出了描述规模报酬不变和规模报酬可变假设的DEA模型及其对偶规划的一种形式。模型克服传统DEA模型应用中常见的决策单元权重为零进而高估决策单元效率的不合理现象,有效地测算了中国31个地区纺织工业的经济运行效率,并分析了地区间效率水平差异的原因。利用各决策单元在经验生产前沿面上的投影点,估计了中国纺织工业的随机生产前沿函数,最后用效率弹性线性递减模型分析了销售收入、人均资产、台港澳和外商投资对各地区效率的微观影响关系。     

Measuring human development via data envelopment analysis: the case of Asia and the Pacific

The United Nations Development Program introduced in 1990 the Human Development Index (HDI), which ever since is published annually in the Human Development Report. In this paper, the assessment of the HDI is reconsidered in the light of data envelopment analysis (DEA). The new approach is applied to the countries of the regional aggregate of Asia and the Pacific. Human development is benchmarked on the basis of empirical observations of best practice countries. First, on the same line with HDI, we develop a DEA-like linear programming model to assess the relative performance of the countries in terms of human development. Then we extend our calculations with a post-DEA goal-programming model to derive estimates of a new development index by using optimal weights for the socioeconomic indicators. Finally, we estimate the relative efficiency of the countries in converting income to human development. We develop and solve for this purpose a pure DEA model with variable returns to scale.     

Production efficiency in Indian agriculture: An assessment of the post green revolution years

In this paper we use the nonparametric method of Data Envelopment Analysis (DEA) to obtain Pareto-Koopmans measures of technical efficiency of individual states over the years 1970–71 through 2000–01 in a multi-output, multi-input model of agricultural production. We disaggregate overall efficiency into two distinct components representing output and input efficiencies and identify the contribution of individual outputs and inputs to the measured level of overall efficiency. Because introduction of modern inputs has been a major component of the process of modernization of Indian agriculture, we examine to what extent different states succeeded in utilizing the modern inputs compared to the traditional inputs. Variations in the DEA efficiency scores across states and over years is explained in terms of differences in various institutional and demographic factors in a second stage regression analysis.     

Target intermediate products setting in a two-stage system with fairness concern

In a two-stage system with two divisions connected in series, fairly setting the target outputs for the first stage or equivalently the target inputs for the second stage is critical, in order to ensure that the two stages have incentives to collaborate with each other to achieve the best performance of the whole system. Data envelopment analysis (DEA) as a non-parametric approach for efficiency evaluation of multi-input, multi-output systems has drawn a lot of attention. Recently, many two-stage DEA models were developed for studying the internal structures of two-stage systems. However, there was no work studying fair setting of the target intermediate products (or intermediate measures) although unreasonable setting will result in unfairness to the two stages because setting higher (fewer) intermediate measures means that the first (second) stage must make more efforts to achieve the overall production plan. In this paper, a new DEA model taking account of fairness in the setting of the intermediate products is proposed, where the fairness is interpreted based on Nash bargaining game model, in which the two stages negotiate their target efficiencies in the two-stage system based on their individual efficiencies. This approach is illustrated by an empirical application to insurance companies.     

基于最近距离投影的DEA两阶段效率评价方法研究

数据包络分析是进行效率评价最重要的方法之一。传统的数据包络分析理论主要寻找有效前沿面上的最远距离投影，在极大化无效性指数的同时也面临着效率改进的巨大难度和高额成本。对于具有两阶段内部生产结构的决策单元，本文从考虑最小改进难度的视角出发，提出了最近距离投影的两阶段效率评价方法。该方法首先得到所有强有效决策单元的线性组合，且这些组合均占优于被评价的两阶段决策单元。然后建立了两阶段范围调整效率评价模型，在确定具有最近投影距离的占优组合的同时，得到了两阶段评价效率。最后，本文运用我国32家上市银行的年度数据对所提出方法进行了应用验证。