Principles and Practice of Structural Equation Modeling

"Structural equation modelling is another set of statistical models that are very popular in the social sciences. They are often used by people who want to investigate how different kinds of abilities, often called ‘constructs’, relate to one another. People design studies where they give survey instruments, for example, or educational assessments. They then create scores from these. They relate all of these to one another. Essentially, it is a very nice way of taking graphical representations of these relationships, and taking data and quantifying how strong these relationships are. Which variable predicts which other variable? Are there moderating or mediating effects between variables that might influence that relationship? How strong is it? Which direction does it go in? And so on and so forth. The reason why I chose this book in particular is because it is reflective of a series that I have really come to like. In this series, the publisher is really trying to break down relatively complex information around assessment methods for people who are educated but not yet experts. If you compare that with the item response theory handbook, it is much more accessible and much more at an introductory level. Get the weekly Five Books newsletter I think this is such an important kind of work to do in our field. It is the kind of work I identify myself with very much. It is what I call ‘handholding for smart people’. It is the same style in which I co-wrote a book with two colleagues a few years back, and with which Jackie and I have edited our latest handbook. You try to describe the key ideas, the key principles, the key practices in an area at a level where you use technical terms sometimes as well as mathematical equations and graphics but you still talk that all through, step-by-step, so that you do not lose all of the nuance and abstract it so much that you trivialise the ideas. I think sometimes that colleagues who are scientists think that that is maybe not as valuable, and it is much more valuable to produce very technical publications in peer review journals, but I personally think this kind of book is a very important contribution. It turns out that this topic, structural equation modelling, represents a very popular, very important family of models. This particular book is already in its fourth edition, so it has clearly found a lot of people who appreciate it practically. Imagine you have an application where you are looking at the relationship between different competencies in English language. Let’s say you have three variables: writing competency, speaking competency, and interpersonal communicative competence. You are interested in how these relate to background variables that people bring to bear in assessment. Maybe the kind of educational background that they have, the kinds of households that they come from, or the educational context in which they are learning English. You might also be interested in how certain kinds of non-cognitive factors like motivation, grit, or persistence mediate how they use these competencies to solve tasks. With structural equation modelling you can set up a model where you have these different constructs represented and you can try to see, say, whether one is predictive of the other."