Book Review: Beaujean, A.A. (2014). Latent Variable Modeling Using R. Hove: Routledge.

LaVaMo_RThis is one of the many stats books I’ve been ordering for the Goldsmiths Library lately (and, boy, I really love ordering new stats books for the library!). Just beware, there are more review of advanced stats books to come over future weeks….

What is R?

This one is an introduction to latent variable modelling and makes use of the R package lavaan which stands for LAtent VAriable ANalysis.  In case you haven’t  heard of it yet, R is a software environment / language / program that is free and open source and offers thousands of packages for all the different types of statistical analysis that anyone could ever think of. The only downside of R over other software programs that psychologists commonly use, is that you can’t get your analysis done by clicking and pointing on a graphical user interface like you would do with SPSS for example. Instead, in R you need to type commands to get anything done, and of course it takes a little while to learn those commands and understand what they are doing. But once you get over this initial hurdle the reward is that the complete world of contemporary statistical analysis procedures is open and free to you and you’ll never have to face that shock again when your software license runs out at the end of July and you haven’t finished your project yet.

Anyway, in R there are several packages for latent variable analysis (or structural equation modelling which is another name for it), but lavaan seems to be the most popular one at the moment. This is probably because it is quite flexible in terms of what kind of data you can analyse with it but much less complicated to use than an alternative package, openMX which is ultra-flexible regarding analysis options and favoured by, for example, the behavioural genetics community but also has a very steep learning curve, even after you think you’ve mastered R.

What is Latent Variable Modelling? 

To be honest, Beaujean’s book isn’t a psychology book per se but latent variable models are of truly high importance to psychologists.  Why is this? Because almost by definition, most questions that psychologists are interested in, involve things we can’t really observe directly with our senses. I’m referring to thinks like intelligence (see below) autism, personality traits, musicality, cognitive deficits, happiness ….. But, even though you can’t observe intelligence or autism directly you might be able to infer from their behaviour whether someone is autistic or not or whether they’re high or low on intelligence. If you are really clever, and dedicated to the question, you might even develop a test or a diagnostic battery which helps you gather observations and data to make inference about the latent construct (intelligence, autism ….) that you are actually interested in. And these are the latent variables that latent variable analysis is all about. So, if we were honest we would have to acknowledge that most models in psychology are actually latent variable models.

Beaujean’s text  
Broadly speaking, Beaujean’s book serves two purposes: First of all, it introduces the main concepts and variants of latent variable analysis (including path analysis, factor analysis, structural equation modelling, latent growth curves, item response models, hierarchical latent models) that you are most likely want to use at some point if you are a working in psychology or educational research. But secondly, after introducing each type of model in very concise terms, Beaujean also shows you how to perform the corresponding statistical analysis using the commands from the lavaan package.

The amazing thing about the book is how clearly it is written both in terms of explaining advanced statistics and in terms of teaching you how to run those analyses yourself. This clarity in writing and its educational mission to really empower the reader to construct their own latent variable models, distinguishes the book from other excellent textbooks on structural equation or latent variable models that either use a lot of maths or don’t cover the breadth of latent variable models that Beaujean is able to present within the 150 pages of this book. Actually, 150 pages is not quite true because the book also contains an extremely useful appendix of about 50 pages that discusses all the corey issues of things like different model fit indices and a glossary that would have cluttered the main text. The appendix also has the answers to the thought-through exercises that Beaujean gives at the end of each chapter. If you’ve got the time to run through at least some of these then you can learn quite a lot about latent variable modelling  – try it out!

Daniel Müllensiefen
This entry was posted in Uncategorized. Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s