My colleague Doug recently completed a research study on the level of engagement in the Federal workforce. He used data from a 2005 governmentwide survey of nearly 37,000 Federal employees. Although Doug has a solid background in Federal employment issues like employee engagement, this was the first time he used a group of survey questions to create and analyze a measurement scale. This book is one of the resources Doug used to make his project a success.
Robert DeVellis's book covers the fundamentals of social science scale development in a straightforward manner. This book explains basic measurement concepts clearly and contains sufficient practical guidance to support construction of a working scale. The reader will need to obtain access to a statistical program and instruction in its use from another source.
Chapter 1 briefly reviews the history of social science measurement, including the role played by statistics and psychophysics. A discussion of the relationship between theory and measurement includes the risks of careless measurement practice. It ends on page 13 with a useful one-paragraph preview of the remaining seven chapters. Chapter 2 defines the relationship between constructs and the measures that allow us to observe them. It introduces path diagrams and outlines the assumptions of classical measurement theory. Chapter 3 defines measurement reliability and introduces coefficient alpha as a measure of the internal consistency of a scale. More advanced reliability topics are outlined with some reference to formulas and covariance matrices.
The next two chapters are the book's core. Chapter 4 defines content, criterion-related, construct and face validity and distinguishes between validity and accuracy. The discussion of validity coefficients and multi-method multi-trait approaches to studying validity equip the reader to understand validity studies in the measurement literature. Chapter 5 lays out an eight-step process for developing a scale of questions to measure some construct of the reader's choice. These steps are (slightly reworded):
- 1. Define clearly what you want to measure.
- 2. Create a set of draft questions.
- 3. Select a common format and set of answer options for the questions.
- 4. Have experts review and revise the questions.
- 5. Consider using "social desirability" or similar questions.
- 6. Field test the questions with "real people."
- 7. Analyze the results of your field test.
- 8. Decide how many questions--and which questions--to keep.
The real value in this book is the practical guidance given for each of these steps. There is enough here to get you through your first project, but not so much that it overwhelms.
Chapter 6 introduces factor analysis as a statistical procedure that helps scale developers understand how their scale works, particularly if there are two or three different things that the scale is measuring. The author does an excellent job explaining the concepts of factor analysis, how to select the right kind of analysis, and how to interpret the results without becoming mired in unnecessary technical detail. Chapter 7 is a similarly elegant treatment of item response theory. The reader is convinced that scale items each have a certain difficulty for test takers and a certain ability to discriminate between groups of test takers. The chapter explains these and related concepts sufficiently to illustrate their usefulness, leaving interested readers to learn more in one of the cited references. The final chapter encourages readers to take a broad, contextual view of measurement and sends them on their way to develop their scales.
Other sources are a better choice if you need a deeply technical reference about measurement (Psychometric Theory), factor analysis (Latent Variable Models and Factor Analysis), or item response theory (Item Response Theory for Psychologists). If you are developing your first scale--like Doug--or if you are often asked "How do I make a scale to measure this?" then you want this book close at hand.