Regression Inside Out

Demonstrates new ways to extract knowledge from statistical data and unlock more nuanced interpretations than has previously been possible.

Eric W. Schoon (Author), David Melamed (Author), Ronald L. Breiger (Author)

9781108841108, Cambridge University Press

Hardback, published 22 February 2024

282 pages
25.1 x 17.6 x 2.1 cm, 0.654 kg

'[T]his text combines ingenuous ideas, uncompromising rigor, and intuitive accessibility. I highly recommend this book to data analysts and statistical practitioners. Its innovative approach and rigorous analysis offer valuable insights that can transform how we interpret regression models.' Jun Xu, Social Forces

Linear regression analysis, with its many generalizations, is the predominant quantitative method used throughout the social sciences and beyond. The goal of the method is to study relations among variables. In this book, Schoon, Melamed and Breiger turn regression modeling inside out to put the emphasis on the cases (people, organizations, and nations) that comprise the variables. By re-analyzing influential published research, they reveal new insights and present a principled way to unlock a set of more nuanced interpretations than has previously been attainable. The emphasis is on intuition and examples that can be reproduced using the code and datasets provided. Relating their contributions to methodologies that operate under quite different philosophical assumptions, the authors advance multi-method social science and help to bridge the divide between quantitative and qualitative research. The result is a modern, accessible, and innovative take on extracting knowledge from data.

1. Regression inside out
Part I: 2. OLS inside out
3. Generalizing regression inside out
4. Turning variance inside out with Eunsung Yoon
Part II: 5. Action detection
6. Interaction detection
Part III: 7. RIO as a gateway to case selection
8. RIO as a gateway to configurational comparative analysis
9. RIO as a gateway to field theory
10. Conclusion
Appendix A: A brief introduction to matrices and matrix multiplication
Appendix B: Computation of the singular value decomposition (SVD)
Appendix C: Variance for binomial and count outcomes
Appendix D: Compositional effects in using RIO to detect statistical interactions
Appendix E: Monte Carlo simulation detecting interactions by regressing on rows of P.

Subject Areas: Social research & statistics [JHBC]