Book Reviews

Abstract

Imagine, that you are a beginner Mplus user and you are left alone with your database and Mplus late at night in your office. Frustrating. Scary. Terrifying. Horrific! What you need is a freshly brewed cup of coffee and Christian Geiser’s book, Data analysis with Mplus. Success is guaranteed. He can help you do the first steps quickly, systematically and professionally in a world full of Mplus mysteries. And of course not only are the first steps guaranteed to succeed, but the very final ones as well. As Kline (2005) describes the process very adequately: “Learning to use and understand a new set of statistical procedures is like making a long journey through a strange land. Such a journey requires a substantial commitment of time, patience, and willingness to tolerate the frustration of some initial uncertainty and inevitable trial and error. But this is one journey that you do not have to make alone.” Geiser is someone who can accompany you on the road, who can make you understand what the particular syntax means, someone who can teach you useful tricks never seen before. As he writes in the preface of the book “numerous screen shots and output excerpts guide the reader through analyses and the interpretation of the Mplus output in a step-by-step fashion”. These analyses are the most fundamental and common ones that an applied (modern) researcher ever needs: linear structural equation models (i.e. simple linear regression analysis, latent regression analysis, confirmatory factor analysis, path models, and mediation analysis); SEM models for measuring variability and change (latent state analysis, autoregressive models, latent growth curve models); multilevel regression analysis and latent class analysis. With the directions of the author, you will have an idea how to model complex relationships between continuous variables at the latent level, you will be able to separate stable from time-specific influences on psychological measures in a longitudinal dataset. You will also be able to avoid biased results (underestimation of standard errors) if you have clustered or nested data structure (i.e. students in classes and schools), and you will have the superpower to classify individuals into homoge neous subgroups (latent classes, latent types) even if you have not seen Mplus before. The step-by-step fashion really is what it means literally. For every topic you can find a brief, but meaningful theoretical introduction that discusses the goals and the situation-specific effects of the given type of analysis including its shortcomings. The next step is the illustration of the analysis in the simplest form possible. At this stage the author applies path diagrams (instead of mathematical equations) that help the reader visualize the outlined relationships and associations within the particular model. The next step is the specification of a simple model (what could be easily generalized to a more complicated level), and the explanation of the input commands. After running your model you can see what you have done. If you are a well-trained Mplus user, you know that there are magic words for which we are praying, namely: “The model estimation terminated normally”. This book can give you the key to the relief of reading this sentence on your screen. It often happens – according to my several years of experience – that it is hard to generalize Mplus User’s Guide’s (Muthén & Muthén, 1998–2012) examples to a particular problem of interest. The User’s Guide is not a prolix one. One can easily end up making hundreds of trivial mistakes, just because mere mortal human beings are prone to do so. Data analysis with Mplus will help to avoid these trivial mistakes. The notations of the most important parameters that are estimated in SEMs are the most common ones. In this book the author introduces these parameters in a straightforward fashion, which makes the text easy to follow and acquire. “Boxes” within each topic are especially useful, because they consist of summaries of the core characteristics and special features of the particular analysis. At times the boxes also provide examples to practice. I might have hinted above that this book is a beginners’ tutorial. Well, this is not entirely true. Actually, this book can help you at any stage of your professional career. You can use the book as a self-study guide and practice with the examples discussed in it. These datasets are all original ones, which can be downloaded from the companion website with the related input and output files (http://crmda.ku.edu/guilford/geiser). Christian Geiser, PhD, is an excellent teacher. In reality he is Assistant Professor at the Department of Psychology at Utah State University in Logan. He has the skills to bridge the gap between theory and practice, and he can translate and transform the meaning into palatable pieces. I had the impression that he knew exactly (by some extraterrestrial skills) where I had blanks in my knowledge. Farsightedly, he made those blank and confused spots disappear. Warning! The book can give rise to enlightened moments!