This volume is a companion and update to Event History Analysis and substantially extends the practical application of event history analysis. It also adds several important new models and concepts which have been developed in an extremely active research area since the late 1980s. It provides a comprehensive introductory account of event history modeling techniques and their usefulness for causal analysis in the social sciences. By giving many concrete application examples, it demonstrates that event history models allow a natural time-related representation of causal arguments in empirical studies. In contrast to structural equation analysis, which is based upon the observation of states and on "time-less" models, event history analysis employs the time-path of changes in states and relates changes in (qualitative and metric) causal variables in the past to changes in discrete outcomes in the future. Since effects follow their causes in time, this necessarily implies a temporal interval which may be very short or very long, but can never be zero or infinity. This book demonstrates that event history modeling is a major step forward in causal analysis because it is the most appropriate of all currently available methodologies to uncover such (normally unknown) lags between causes and their effects and reveals the various temporal shapes of the unfolding effect. This is because the transition rate can be used to represent the quantity of the causal effect at any point in time. A particular strength of this book lies in the description of a new approach to interdependent dynamical systems. It is shown that a causal approach to interdependent systems is easily possible with the help of the transition rate concept, and that the systems view is not a substitute for a proper causal approach in the social sciences.
This book also proposes that the social sciences should give up their traditional deterministic approach in empirical analyses in favor of a probabilistic one. It is argued that randomness should not only be seen as a technical term (arising because of limited empirical observation), but must be understood as a theoretical category; it is the propensity of social agents to change their behavior in the future under certain conditions that have taken place in the past and present. This means that the aim of statistical (and substantive) models must be to capture common elements in the behavior of people, or patterns of action that recur in many cases. In event history models, the causal effect to be explained is therefore the probability of a time-related change. Finally, the book is critical with regard to the widely applied models with unobserved heterogeneity since there is, in general, no way to make reliable assumptions about what has not been observed. Thus, in using such models, most empirical researchers try to draw sharp conclusions, even when these can only be generated by imposing much stronger assumptions than can be plausibly defended.
This book introduces the reader to the computer program TDA (Transition Data Analysis). Designed by Götz Rohwer, TDA estimates the kinds of models most frequently used with longitudinal data, in particular, event history data. The guiding principle in constructing TDA was the desire to make a broad range of event history analysis techniques as simple and convenient to apply as possible. TDA is now widely used in many research and university centers which analyze longitudinal data in Europe and the USA. It can be run on DOS-based personal computers and UNIX workstations. Included with this book is a disk with an executable version of the TDA program package for DOS-based machines, a file with the data used in the examples throughout the book, and a series of files containing the TDA set-ups for the examples. Thus, the reader is offered the unique opportunity to easily run and modify all the application examples on the computer. The authors have emphasized the strengths and weaknesses of event history modeling techniques in each example. In particular, they have complemented each practical application with a short exposition of the underlying statistical concepts. The examples start with an introduction to the background for the specific model. Then there is a demonstration of how to organize the input data and use the commands to control the TDA operation. Finally, a substantive interpretation of the obtained results is given.
