Practical Longitudinal Data Analysis

Our previous book on repeated measures data (Crowder and Hand, 1990) ranged fairly widely, bringing together a large and scattered literature on the subject. The present book originally began as a second edition to that. However, recognizing that several other books on the subject had appeared since 1990 (section 1.5), we decided instead to produce something more restricted: we decided to concentrate on regression-based models. Thus, although this book has much overlap with our previous one, it is intended to complement it: there are fewer topics here, but they are covered in more depth and with more discussion. Particular features of this book are: r-We emphasize statistical, rather than probabilistic, models. That is, models are regarded as adequate vehicles for inference about the general behaviour of the observations rather than as true descriptions of the underlying stochastic mechanisms responsible for generating the data. -The text is illustrated with simple numerical examples called illustrations and most chapters contain larger real examples. The latter have been analyzed is available (it does not yet exist for all using commercial software when this of the methods described in the book), and illustrative computer commands and results are given. This is one reason why we have used the word 'practical' in the title. -The first half of the book deals with normal regression models and the second half with increasingly non-normal data. We handle non-normal distribu­ tions by fitting normal type models-another reason for adopting the word 'practical'.