Techniques of Event History Modeling: New Approaches to Causal Analysis

Techniques of Event History Modeling: New Approaches to Causal Analysis

Techniques of Event History Modeling: New Approaches to Causal Analysis

Techniques of Event History Modeling: New Approaches to Causal Analysis

Synopsis

Including new developments and publications which have appeared since the publication of the first edition in 1995, this second edition: *gives a comprehensive introductory account of event history modeling techniques and their use in applied research in economics and the social sciences; *demonstrates that event history modeling is a major step forward in causal analysis. To do so the authors show that event history models employ the time-path of changes in states and relate changes in causal variables in the past to changes in discrete outcomes in the future; and *introduces the reader to the computer program Transition Data Analysis (TDA). This software estimates the sort of models most frequently used with longitudinal data, in particular, discrete-time and continuous-time event history data. Techniques of Event History Modeling can serve as a student textbook in the fields of statistics, economics, the social sciences, psychology, and the political sciences. It can also be used as a reference for scientists in all fields of research.

Excerpt

This chapter discusses event history data structures. We first introduce the basic terminology used for event history data and then give an Example of an event history data file. Finally, we show how to use it with TDA.

2.1 Basic Terminology

Event history analysis studies transitions across a set of discrete states, including the length of time intervals between entry to and exit from specific states. The basic analytical frimework is a state space and a time axis. The choice of the time axis or clock (e.g. age, experience, marriage duration, etc.) used in the analysis must be based on theoretical considerations and affects the statistical model. In this book, we discuss only methods and models using a continuous time axis. An episode, spell, waiting time, or duration—terms that are used interchangeably—is the time span a unit of analysis (e.g. an individual) spends in a specific state. The states are discrete and usually small in number. The definition of a set of possible states, called the state space Ɣ;, is also dependent on substantive considerations. Thus, a careful, theoretically driven choice of the time axis and design of state space are important because they are often serious sources of misspecification. In particular, misspecification of the model may occur because some of the important states are not observed. For Example, in a study analyzing the determinants of women's labor market participation in West Germany, Blossfeld and Rohwer (1997) have shown that one arrives at much more appropriate substantive conclusions if one differentiates the state “employed” into “full-time work” and “part-time work.” One should also note here that a small change in the focus of the substantive issue in question, leading to a new definition of the state space, often requires a fundimental reorganization of the event history data file.

The most restricted event history model is based on a process with only a single episode and two states (one origin and one destination state). An Example may be the duration of first marriage until the end of the marriage, for whatever reason. In this case each individual who entered into first marriage (origin state) started an episode, which could be terminated by a transition to the destination state “not married anymore.” In the single episode case each unit of analysis that entered into the origin state is represented by one episode. If more than one destination state exists, we . . .

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.