理论与方法专栏 | 语言学数据的回归分析

编者按

《理论语言学五道口站》（2023年第28期，总第292期）理论与方法专栏为大家推荐加拿大麦吉尔大学语言学系副教授Morgan Sonderegger的著作：Regression Modeling for Linguistic Data（《语言学数据的回归分析》）。该书已于2023年6月在MIT Press出版。

著作简介

 Regression Modeling for Linguistic Data

《语言学数据的回归分析》

Abstract：

In the first comprehensive textbook on regression modeling for linguistic data in a frequentist framework, Morgan Sonderegger provides graduate students and researchers with an incisive conceptual overview along with worked examples that teach practical skills for realistic data analysis. The book features extensive treatment of mixed-effects regression models, the most widely used statistical method for analyzing linguistic data.

Sonderegger begins with preliminaries to regression modeling: assumptions, inferential statistics, hypothesis testing, power, and other errors. He then covers regression models for non-clustered data: linear regression, model selection and validation, logistic regression, and applied topics such as contrast coding and nonlinear effects. The last three chapters discuss regression models for clustered data: linear and logistic mixed-effects models as well as model predictions, convergence, and model selection. The book's focused scope and practical emphasis will equip readers to implement these methods and understand how they are used in current work.

简介：

该书是频率主义框架下第一本语言学数据回归分析的综合型教程。书中，作者Morgan Sonderegger为相关专业的研究生和学者提供了深入的概念讲解，并结合工作实例对数据分析的实战技巧进行了教学。该书的特点在于大量运用了混合效应回归模型，而这也是语言学数据分析中应用最广泛的统计方法。

Morgan Sonderegger的讲解以回归分析的基础知识为起点，从假设开始，到推断统计，再到假设检验、统计效力以及其他误差。随后，作者还介绍了非聚类数据的回归分析，包括线性回归、模型选择及验证、逻辑回归和一些应用层面的讨论，如对照编码和非线性效应。该书的最后三章讨论了聚类数据的回归模型，包括线性混合效应模型、逻辑混合效应模型以及模型的预测、收敛和选择。该书侧重明确，注重实操，能够让读者很好地掌握书中的统计方法，并理解这些方法在实际研究中的工作原理。

作者简介

-Morgan Sonderegger-

Morgan Sonderegger，加拿大麦吉尔大学语言学系副教授、蒙特利尔计算语言学与定量语言学实验室（Montréal Computational & Quantitative Linguistics Lab，MCQLL）联合负责人。研究领域包括音系学、语音学、计算语言学、语言变异与变迁、定量研究方法等。曾于2017至2020年期间主持加拿大地区的英语方言话语调查（Speech Across Dialects of English，SPADE）项目。

Morgan Sonderegger is Associate Professor in the Department of Linguistics at McGill University and co-PI of Montréal Computational & Quantitative Linguistics Lab (MCQLL). His research interests include phonology, phonetics, computational linguistics, language variation and change, quantitative methods, etc. He was the Canada PI for the Speech Across Dialects of English (SPADE) project during 2017 to 2020.

本文版权归“理论语言学五道口站”所有，转载请联系本平台。

编辑：董泽扬 黄静雯

排版：董泽扬 黄静雯

审校：吴伟韬 时仲

英文编审负责人：董泽扬