A Review of Multilevel and Longitudinal Modeling Using Stata

In their new edition, Rabe-Hesketh and Skrondal add to the maturing applied literature on multilevel modeling techniques with an in-depth demonstration in Stata. Suitable for self-study or an applied first graduate course, the book goes well beyond simply covering the relevant command syntax. Starting with a review of simple regression, it presents a largely nonspecialist account of the most common multilevel extensions. Repeated illustrations are also included of many software features—descriptive statistics, hypothesis tests, data recoding and restructuring, basic data simulations, and custom graphing—of practical interest beyond the modeling techniques at hand. Every chapter includes study exercises, and sample data used in the examples can be retrieved directly into Stata from the Internet. Most formulas and notation will be understood by any reader with a reasonable grasp of linear regression. Each topic is developed through case studies of one or more data sets and includes discussion of inference and interpretation of results. A few tips on model identification are included throughout, although it is not a major focus. A chapter is dedicated to panel data and estimating growth curves, while another outlines crossed random-effects models in contrast to their more familiar nested-effects cousins. Treatment of longitudinal data has been expanded considerably since the first edition. The authors also provide alternate equation notation in several places to ease the transition between the competing multilevel notation schemes, and they helpfully point out a few inconsistencies in notation between various Stata procedures. The book is geared toward Stata version 10, but most commands will work in other recent versions. The book’s strengths include diverse data examples drawn from disciplines ranging from medicine and public health to education and economics. It displays a generous amount of command syntax and procedure output, which makes it easy to follow and replicate its sample analyses. Another high point is a section