以缺失的数据组合:哪些集束分析方法的估算模型? (Clustering with missing data: which imputation model for which cluster analysis method?)

Multiple imputation (MI) is a popular method for dealing with missing values. One main advantage of MI is to separate the imputation phase and the analysis one. However, both are related since they are based on distribution assumptions that have to be consistent. This point is well known as congeniality. In this paper, we discuss congeniality for clustering on continuous data. First, we theoretically highlight how two joint modeling (JM) MI methods (JM-GL and JM-DP) are congenial with various clustering methods. Then, we propose a new fully conditional specification (FCS) MI method with the same theoretical properties as JM-GL. Finally, we extend this FCS MI method to account for more complex distributions. Based on an extensive simulation study, all MI methods are compared for various cluster analysis methods (k-means, k-medoids, mixture model, hierarchical clustering). This study highlights the partition accuracy is improved when the imputation model accounts for clustered individuals. From this point of view, standard MI methods ignoring such a structure should be avoided. JM-GL and JM-DP should be recommended when data are distributed according to a gaussian mixture model, while FCS methods outperform JM ones on more complex data.

翻译：多重估算(MI) 是处理缺失值的流行方法。 MI的主要优势之一是将估算阶段与分析阶段区分开来。但是, 两者都相关, 因为它们基于分布假设, 并且必须保持一致。这一点众所周知。在本文中, 我们讨论对连续数据分组的共性。首先, 我们理论上强调两种联合模型( JM- GL 和 JM- DP ) 方法( JM- GL 和 JM- DP ) 是如何与各种分组方法相匹配的。然后, 我们提出一个新的完全有条件的 MI 方法( FCS ), 与 JM- GL 具有相同的理论属性。最后, 我们扩展了 FCS MI 方法, 以核算更为复杂的分布。在广泛的模拟研究中, 所有MI 方法都比较了各种群集分析方法( k- 平均值、 k- mid 、混合模型、等级组合组合组合 ) 。这项研究强调, 当对聚类集个人进行估算模型核算时, 分区精确度的精确度将会提高。从这个结构的标准的MI法方法应该避免。 JM- L 和JM- DMIS 数据在模型中进行。