Books by William D. Berry

Where an assumption of unidirectionality in causal effects is unrealistic, 'recursive' models cannot be used, and complex 'nonrecursive' models are necessary. But, many nonrecursive models are 'unidentified', which makes meaningful parameter estimation impossible. This book explains the concept of identification and the factors that lead to it.

Through the use of careful explanation and examples, Berry demonstrates how to consider whether the assumptions of multiple regression are actually satisfied in a particular research project. Beginning with a brief review of the regression assumptions as they are typically presented in text books, he moves on to explore in detail the substantive meaning of each assumption, for example, lack of measurement error, absence of specification error, linearity, homoscedasticity and lack of autocorrelation.