This is a book-- as the authors state-- that give the reader an understanding of multivariate statistics in a way analogous to the (usual, mainstream) driver of automobiles-- KNOWING HOW TO USE THE EQUIPAGE WITHOUT BY NECESSITY BEING SUPER-TECH!!! In other words, one can by thoughtful reading of this book get the "gist" of multivariate statistics without the ponderous math lying beneath this type of software-- such as SPSS (all versions) SAS (all versions) etc. etc. Understanding of this sort will help one drive-through-the-maze of software applications-- all of which will do SOME multivariate computation-- but some much more fittingly for one's purpose.
I find it to be true that many in social science research will simply resort to guess-ology when it comes to this sort of quantitative work: they will go from data-input to a supposed multivariate computation without such understanding as Tabachnick and Fidell could provide-- and then blindly just-read the output. This may come close to exactly being garbage-in-garbage-out (GIGO).
What I am positing therefore is that one needs to have a sense of the logic of software in any event before unquestioningly inputting data, and "reading." If one does not have such an ground-up understanding, the mistakes that can be made-- and assumed to be correct-- abound. But if one has a solid understanding of the presentation in Tabachnick and Fidell, this is the-less-likely to occur.
The authoresses begin with an overview-- in a few paragraphs per type of program deriving from a decision-tree-diagram for all multivariate software. Then separate chapters for each type of software are provided, so that one starts from the general and goes to the very-specific-indeed. By a careful analysis of these chapters and overview, one can avoid the illusory/mistake-rife tendency to "just put in the data and see what comes out."
In short, this is just about the best overview-book on multivariate stats on the market, and has been so for some time (decades, really.) All that is pre-required would be a good grasp of univariate (STATS 101) statistics, and a good head for sensible-research-questioning. If one's design is good/valid-- and if one has done the fact-finding homework-- the product will augment the result of univariate work by several orders of magnitide.
So: by all means: get this book, empirically-minded researchers! Since all things are headed in a multivariate cybernetic direction, this acquisition will never be mistaken!
