Get 20M+ Full-Text Papers For Less Than $1.50/day. Subscribe now for You or Your Team.

Learn More →

Multiple Regression: Testing and Interpreting Interactions

Multiple Regression: Testing and Interpreting Interactions The Statistician (1994) 43, No.3. pp. 453--473 Book Reviews L. S. AIKEN and S. G. WEST, 1991 Newbury Park, Sage xii + 212 pp. ISBN 0 8039 3605 2 The authors of this book aim to provide social scientists with clear prescriptions for the probing and interpretation of continuous variable interactions that are analogues of existing prescriptions for categorical variable interactions. They criticize the underutilization of multiple regression and the use of procedures such as median splitting followed by the analysis of variance. However, their adoption and use of multiple regression is uncritical. There is little discussion on or examination of the appropriateness of the assumptions on which multiple regression is based and no mention of other approaches such as generalized linear models. The examples given are abstract or artificial, and there is little evidence of the modelling that should precede a scientific investigation and that links application to theory. However, if multiple regression can be applied as a black box procedure in the social sciences, does this book meet the requirements of the target audience of researchers and graduate students who are already familiar with multiple-regression analysis? The contents are summarized by a list of chapter titles: 'Interactions between continuous predictors in multiple regression', 'The effects of predictor scaling on coefficients of regression equations', 'Testing and probing three-way interactions', 'Structur­ ing regression equations to reflect higher order relationships', 'Model and effect testing with higher order terms', 'Interactions between categorical and continuous variables', 'Reliability and statistical power' and 'Conclusion: some contrasts between ANOVA and MR in practice'. Matrix notation is not used except in optional sections and the models generally involve only two predictor variables with first-order and some second-order terms. Particular models are examined discoursively. Results are often quoted without derivation and thus are difficult to check. The authors have many worthy aims, e.g. to increase awareness of the importance of identifying possible interactions and of methods for exploring their significance, and the need to recognize what is genuine and what is the consequence of the scale of measurement used. However, the approach adopted which is common in the social sciences, of attempting to present ideas based on complex mathematics without using mathematics, leads to a difficult discoursive presentation that is equally unlikely to be understood by the target audience. Perhaps it is time for researchers in the numerical social sciences, as is normal for their contemporaries in the numerical sciences, to be expected to and be trained to understand the statistical theory underpinning the methods and sofware that they use. Though this book has many worthy aims, it presents a black box approach to applying multiple regression in the social sciences that I cannot support. Thus I cannot recommend it to the target audience. However, I do recommend that those teaching multiple regression to the target audience should read it and include testing and interpreting interactions in their teaching. M. A. PORTER Kirkby-in-Furness Inverse Methods in Physical Oceanography A. F. BENNETT, 1992 Cambridge, Cambridge University Press + 346 pp., £35 xvi ISBN 0 521 385687 Why is an oceanographic book being reviewed in The Statistician? The answer lies in inverse models. Oceanographers use complicated fluid dynamical' models to predict and explain the flow of water (and other parameters such as heat or salt) in the ocean. Inverse models are an attempt to combine such © 1994 Royal Statistical Society 0039-0526/94/43453 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of the Royal Statistical Society Series D: The Statistician Oxford University Press

Multiple Regression: Testing and Interpreting Interactions

Loading next page...
 
/lp/oxford-university-press/multiple-regression-testing-and-interpreting-interactions-G0hzQEO3lF

References (3)

Publisher
Oxford University Press
Copyright
© 1994 Royal Statistical Society
ISSN
2515-7884
eISSN
1467-9884
DOI
10.2307/2348581
Publisher site
See Article on Publisher Site

Abstract

The Statistician (1994) 43, No.3. pp. 453--473 Book Reviews L. S. AIKEN and S. G. WEST, 1991 Newbury Park, Sage xii + 212 pp. ISBN 0 8039 3605 2 The authors of this book aim to provide social scientists with clear prescriptions for the probing and interpretation of continuous variable interactions that are analogues of existing prescriptions for categorical variable interactions. They criticize the underutilization of multiple regression and the use of procedures such as median splitting followed by the analysis of variance. However, their adoption and use of multiple regression is uncritical. There is little discussion on or examination of the appropriateness of the assumptions on which multiple regression is based and no mention of other approaches such as generalized linear models. The examples given are abstract or artificial, and there is little evidence of the modelling that should precede a scientific investigation and that links application to theory. However, if multiple regression can be applied as a black box procedure in the social sciences, does this book meet the requirements of the target audience of researchers and graduate students who are already familiar with multiple-regression analysis? The contents are summarized by a list of chapter titles: 'Interactions between continuous predictors in multiple regression', 'The effects of predictor scaling on coefficients of regression equations', 'Testing and probing three-way interactions', 'Structur­ ing regression equations to reflect higher order relationships', 'Model and effect testing with higher order terms', 'Interactions between categorical and continuous variables', 'Reliability and statistical power' and 'Conclusion: some contrasts between ANOVA and MR in practice'. Matrix notation is not used except in optional sections and the models generally involve only two predictor variables with first-order and some second-order terms. Particular models are examined discoursively. Results are often quoted without derivation and thus are difficult to check. The authors have many worthy aims, e.g. to increase awareness of the importance of identifying possible interactions and of methods for exploring their significance, and the need to recognize what is genuine and what is the consequence of the scale of measurement used. However, the approach adopted which is common in the social sciences, of attempting to present ideas based on complex mathematics without using mathematics, leads to a difficult discoursive presentation that is equally unlikely to be understood by the target audience. Perhaps it is time for researchers in the numerical social sciences, as is normal for their contemporaries in the numerical sciences, to be expected to and be trained to understand the statistical theory underpinning the methods and sofware that they use. Though this book has many worthy aims, it presents a black box approach to applying multiple regression in the social sciences that I cannot support. Thus I cannot recommend it to the target audience. However, I do recommend that those teaching multiple regression to the target audience should read it and include testing and interpreting interactions in their teaching. M. A. PORTER Kirkby-in-Furness Inverse Methods in Physical Oceanography A. F. BENNETT, 1992 Cambridge, Cambridge University Press + 346 pp., £35 xvi ISBN 0 521 385687 Why is an oceanographic book being reviewed in The Statistician? The answer lies in inverse models. Oceanographers use complicated fluid dynamical' models to predict and explain the flow of water (and other parameters such as heat or salt) in the ocean. Inverse models are an attempt to combine such © 1994 Royal Statistical Society 0039-0526/94/43453

Journal

Journal of the Royal Statistical Society Series D: The StatisticianOxford University Press

Published: Dec 5, 2018

There are no references for this article.