语言学定量研究中的稳健统计方法

Robust Statistical Methods for Quantitative Research in Linguistics

稳健统计学的进展为研究者探索数据、发现模式和检验研究假设提供了有价值的技术。本文介绍稳健统计学的基本原理和基本概念。举例说明在常规的方差分析统计假设不满足的情况下如何使用Welch型方差分析等几种稳健统计方法,在皮尔逊相关分析统计假设不满足的情况下如何开展百分数弯曲相关和跳跃相关分析。本文建议语言学定量研究者重视常规统计分析假设,学会使用稳健统计方法,提高研究的质量。

Advances in robust statistics provide researchers with valuable techniques to explore data, to discover patterns, and to test research hypotheses. This article introduces the fundamental principles and concepts of robust statistics, followed by an illustration of how to employ several robust statistical methods like a Welch-type ANOVA when the assumptions of a conventional ANOVA are violated, and of how to conduct the percentage bend and skipped correlation analyses when the assumptions of a conventional Pearson's correlation analysis are violated. It is suggested that researchers in quantitative linguistics pay due attention to the assumptions of conventional statistical methods and learn to employ robust statistical methods so as to improve research quality.