LAG SEQUENTIAL-ANALYSIS - ROBUST STATISTICAL-METHODS
LAG SEQUENTIAL-ANALYSIS - ROBUST STATISTICAL-METHODS
复制标题
DOI：
10.1037/0033-2909.101.2.312
复制
发表时间：
1987-03-01
影响因子：
22.4
通讯作者：
DORFMAN, DD
中科院分区：
文献类型：
--
作者：
FARAONE, SV;DORFMAN, DD
腿序贯分析（Sackeh，1979，1980）已成为人际互动研究人员的重要工具。这种方法使人们能够探索复杂的交互行为序列中发生的总结交叉依赖性。滞后序贯分析的统计方法已被发现是不正确的（Allison 和 Liker，1982）或使强有力的理论假设难以验证。在本文中，我们澄清了 Gottman (1979a, 1979b) 和 Sackett (1979) 提出的统计数据的困惑。然后，我们提出了交叉依赖性的统计测试，该测试源自每个行为序列都是一阶马尔可夫链的假设。最后，我们介绍两种稳健的方法：sbd.the jackknife 和 the data split.sbd.，用于测试交叉依赖性并估计交叉依赖性指数的置信区间。我们使用蒙特卡洛模拟来证明这些方法的充分性，并为在它们之间进行选择提供基础。
Leg sequential analysis (Sackeh, 1979, 1980) has become an important tool for researchers of interpersonal interaction. This method enables one to explore the summarize cross-dependencies occurring in complex interactive sequences of behavior. Statistical methods for lag sequential analysis have been found to be incorrect (Allison and Liker, 1982) or make strong theoretical assumptions difficult to verify. In this article we clarify confusion about statistics proposed by Gottman (1979a, 1979b) and Sackett (1979). We then present a statistical test of cross-dependency that derives from the assumption that each behavioral sequence is a first-order Markov chain. Lastly, we introduce two robust methods.sbd.the jackknife and the data split.sbd.for testing cross-dependence and estimating confidence intervals about indices of cross-dependence. We use Monte Carlo simulations to demonstrate the adequacy of these methods and to provide basis for choosing between them.