【作者】 姚婷；

【导师】 陈骥；

【作者基本信息】 浙江工商大学 ， 应用统计硕士（专业学位）， 2021， 硕士

【摘要】 在群组评价中,由于个体在知识经验,学术背景等方面的差异,以及评价对象的复杂性,会导致不同的评价个体之间的差异较大,从而影响评价结果,产生杠杆效应问题。所谓“杠杆效应”,即在群组评价过程中,当出现“舞弊”、对部分评价单元的倾向性偏离、评价能力不稳定等诸多因素时,可能出现歪曲的少数评价个体的评价意见“撬动了”绝大多数正常的评价意见,从而使评价结论偏离多数人的期望和评价对象的实际情况。现有研究中,学者们大多是集中于对群组评价方法、信息集成方法的创新和改进,以及对权重的分配和调整等方面的研究,而较少涉及群组评价中的杠杆效应问题。因此,本文尝试对杠杆效应问题进行了系统的研究,形成了针对杠杆效应的识别测算和处理的方法体系,弥补现有研究对群组评价中的杠杆效应问题的忽视。杠杆效应的识别测算机制。针对群组评价中杠杆效应的识别问题,本文采用“先整体,后局部”的思路,形成了完整的杠杆效应识别测算机制。整体层面,提出了基于个体判断和评价结论的识别方法,同时构建了一个杠杆率指标直接测算杠杆效应水平的高低;个体层面,基于差异度和异常度的思想,从杠杆效应的实施者和承载体两个角度设计了对应的识别测算方法。最后通过应用算例证明了上述方法的可行性。杠杆效应的处理机制。为减弱或消除杠杆效应给群组评价带来的负面影响,本文提出了基于数据不可比的处理机制和基于个体权重分配的处理机制。基于数据不可比的处理机制,本文针对偏向型评价行为和非偏向型评价行为引起的杠杆效应,分别提出了基于OWA算子的数据处理方法和数据标准化处理方法,消除数据的不可比问题,从而弱化杠杆效应的影响。基于个体权重分配的处理机制,首先从静态的角度出发,根据上述的两种评价行为,提出了基于评价表现稳定性的权重分配方法和基于个体差异度的变权调整方法。其次,从动态的角度出发,通过改进多阶段交互反馈群组评价方法,引入杠杆率指标,提出了基于多阶段交互反馈的权重分配方法,通过个体对自身评价信息的调整,逐步消除群组评价中的杠杆效应问题。最后通过应用算例,验证了上述方法能够有效控制不同情况下杠杆效应问题的负面影响。更多还原

【Abstract】 In group evaluation,due to individual differences in knowledge,experience,academic background and other aspects,as well as the complexity of the evaluation objects,there will be a large difference among different evaluation individuals,which will affect the evaluation results and lead to the problem of leverage effect.The so-called "leverage effect",that is,in the process of group evaluation,when there is a "fraud",the orientation deviation of partial evaluation units,evaluation ability is not stable,and many other factors,the possible distortion of a few individual assessment evaluation opinions "leveraging" the vast majority of normal evaluation opinions,so that the evaluation conclusion deviates from most people’s expectations and the actual situation of evaluation objects.In the existing researches,scholars mostly focus on the innovation and improvement of group evaluation methods and information integration methods,as well as the allocation and adjustment of weights,etc.,but seldom involve the problem of leverage effect in group evaluation.Therefore,this paper attempts to conduct a systematic study on the leverage effect,and forms a method system for the identification,measurement and treatment of leverage effect,to make up for the neglect of existing research on the leverage effect in group evaluation.Identification and measurement mechanism of leverage effect.Aiming at the identification problem of leverage effect in group evaluation,this paper adopts the idea of "whole first,then part" to form a complete identification and measurement mechanism of leverage effect.On the overall level,the recognition method based on individual judgment and evaluation conclusion is proposed respectively,and a leverage ratio index is constructed to directly measure the level of leverage effect.At the individual level,based on the idea of difference degree and anomaly degree,the corresponding identification and measurement methods are designed respectively from the two perspectives of the implementer and the carrier of the leverage effect.Finally,the feasibility of the above method is illustrated by an example.The mechanism of the leverage effect.In order to reduce or eliminate the negative impact of leverage effect on group evaluation,this paper proposes a processing mechanism based on data incommensurability and a processing mechanism based on individual weight allocation.Based on the data incommensurability processing mechanism,aiming at the leverage effect caused by the biased evaluation behavior and the non-biased evaluation behavior,this paper proposes the data processing method based on OWA operator and data standardization processing method respectively,to eliminate the data incommensurability problem and thus weaken the influence of leverage effect.Based on the processing mechanism of individual weight allocation,this paper firstly proposes the weight allocation method based on the stability of evaluation performance and the variable weight adjustment method based on the degree of individual difference respectively from the static perspective and according to the above two kinds of evaluation behaviors.Secondly,from a dynamic perspective,based on the improvement of multi-stage interactive feedback group evaluation method and the introduction of leverage ratio index,this paper proposes a weight allocation method based on multi-stage interactive feedback,and gradually eliminates the leverage effect in group evaluation through the adjustment of individual evaluation information.Finally,an example is given to show that the above methods can effectively control the negative effects of leverage under different conditions.更多还原