The benefit of the doubt (BoD) is a non-parametric weighting method that aims to maximize the relative composite indicator value of each decision-making unit (DMU). A well-known issue of BoD-based ranking of DMUs is the rank reversal problem when DMUs are deleted or added. The fundamental reason for rank reversals is that deleting (or adding) a DMU influences the generated weight by changing the frontier surface. To address the rank reversal problem, this paper proposes a subgroup dominance-based BoD model (SD-BoD). Based on the individual weights obtained from a classic BoD model, a dominance-based pairwise comparison mechanism is put forward to ensure comparability of different DMUs. Moreover, motivated by the data envelopment analysis idea, we construct a sequential frontier surface partition technique and establish a DMU ranking framework based on subgroup dominance, which helps reduce weight variation due to DMU addition or deletion. Finally, the effectiveness of the SD-BoD method is illustrated by comparing and analyzing the Human Development Indices of the 28 European regions.

怀疑收益（BoD）是一种非参数加权方法，旨在最大化每个决策单元（DMU）的相对综合指标值。基于 BoD 的 DMU 排名的一个众所周知的问题是删除或添加 DMU 时的排名反转问题。秩反转的根本原因是删除（或添加）DMU 通过改变前沿面来影响生成的权重。为了解决排名反转问题，本文提出了一种基于子群优势的 BoD 模型（SD-BoD）。基于经典BoD模型获得的个体权重，提出了基于优势的成对比较机制，以确保不同DMUs的可比性。此外，受数据包络分析思想的启发，我们构建了顺序边界曲面划分技术，并建立了基于子组优势的 DMU 排序框架，这有助于减少由于 DMU 添加或删除而导致的权重变化。最后，通过比较和分析欧洲28个地区的人类发展指数，说明SD-BoD方法的有效性。