Rank-One Round Robin Designs
提出一类方差分析模型，用于分析轮换互动数据，其中参与者的行动者和伙伴效应线性相关，迫使协方差矩阵秩为1，并给出统计推断和算法。
Abstract A class of analysis of variance models is proposed for analyzing round robin interaction data, where the actor and partner effects of any member are linearly related, thus forcing the covariance matrix of these effects to have rank one. Such data arise frequently in applications, for instance, in sociometric studies on structures of social relationships. A factor analytic approach is proposed for studying these models. Statistical inference about the linear effects is discussed and a convergent algorithm is presented for obtaining the maximum likelihood estimates of the covariance components. A data set is analyzed using the methodology developed.