Usually, multiple regression and causal analysis are treated as separate topics in separate books. McClendon has integrated the two areas within one text, oriented to their application in the social and behavioral sciences. After presenting an overview of issues and techniques for conducting causal analysis, the author devotes six chapters to regression analysis, but from the perspective of causal analysis. The final two chapters unite the two statistical methods, d… More >>
Multiple Regression and Causal Analysis
I purchased this for a stats course in regression. Too simplistic, too short, not enough detail. I purchased a different stats book which did the job at twice the price. Even if this is the text for your course, I recommend finding a different one immediately. It is useful only for review, not for actually learning the material.
Rating: 1 / 5
The author of this book has made a determined effort to write an introduction to regression analysis which will be accessible to readers who want to learn regression but who lack strong backgrounds in mathematical statistics. Folks such as these, and I consider myself one of them, are quite capable of doing substantively interesting and useful research if they are given an opportunity to acquire statistical tools such as multiple regression without first working through proofs and derivations that outstrip their mathematical knowledge. As such, McClendon’s text provides a solid foundation for those who wish to use multiple regression in a creditable way on practical problems of general interest.
This book is not, however, a useful substitute for the kind of text used in a first course in econometrics or regression theory. This limitation, however, should come as no surprise: the author is clear as to his practical, mathematically non-formal purpose from the outset. All tolled, this is a useful text which covers a broader range of material than other books of its type.
I do wish the author had spent less time on non-essential issues, such as the remote possibility that a standardized regression coefficient in a multiple regression equation will not fall between +1 and -1. (In my experience, this happens only in cases of multicollinearity so extreme that the equation is uninterpretable.) There are also easier ways than those laboriously offered by the author to find the contribution to R-squared of any one independent variable: just multiply its bivariate correlation with the dependent variable by its beta weight. In themselves, however, these are not a serious problems, provided that beginners are advised that they are, at most, peripheral issues.
Rating: 4 / 5