也谈差序格局的经验测量：对“厘差定序”法的补充与拓展[1]

胡安宁[2]

摘要：差序格局是理解中国社会人际交往基本模式的重要文化概念。在现有研究中，差序格局的测量可以采用“厘差定序”的方法实现。但是，“厘差定序”法所构建的个体层次的差序程度统计量会因为没有固定个体层次的回归模型截距而产生信度上的偏差。此外，当个体层次的差序程度统计量被用于分析决定个体差序程度的各种因素时，亦存在异方差性的问题。本文通过引入固定值的自我信任选项，解决了个体层面差序程度统计量测量的信度偏差问题。通过展示加权回归模型、自助法和多层次模型三种不同的建模逻辑，详细对比了处理差序程度统计量异方差性问题的多种分析策略。

关键词：差序格局；“厘差定序”法；信度偏差；异方差性

Abstract: The pattern of differential mode of association is a key cultural concept to understand the basic pattern of interpersonal dynamics in Chinese society. In the existing research, the measurement of the differential mode of association can be carried out by the method of "specifying difference and identifying order". However, the individual-level regression model that is adopted in this method incurs a validity problem because there is no fixed individual-level regression intercept. In addition, there is also the problem of heteroscedasticity when the individual-level measure of differential mode of association is subsequently used to analyze various determinants. This paper solves the validity problem of individual-level measure by introducing a fixed-value self-confidence item. In addition, by using three different modeling approaches-weighted regression model, bootstrap method and multi-level model, the strategies to deal with the heteroscedasticity problem in the estimation of the differential mode of association are presented in detail.

Key words: Differential Mode of Association; "Specifying Difference and Identifying Order" Method; Validity Problem; Heteroscedasticity

本周热推：

领导干部法治思维读本养老服务管理陌生人的力量风痕：我与性社会学互构房龙地理