Abstract # 3

Steinhage, Nikolei. (2000). Zeitaggregation und Ereignisdaten. Eine Simulation zu den Auswirkungen der Prozesszeitskalierung. In: Globalife Working Paper Series, No. 03, Faculty of Sociology, University of Bielefeld.

In this paper, Nikolei Steinhage discusses the problem of time-aggregation bias. It often occurs together with another problem called a short-spell problem and is of importance in event-history analysis. The estimation of parameters in event-history analysis normally assumes that time is measured continuously, that events can only occur at discrete points in time, or that the rate between two points in time is constant. These assumptions are seldom fulfilled. Thus, one has to assume that the measurement of events is exact enough to obtain reliable results. The problem is: we only know that the duration of units in a defined state is between T-1 and T. We do not know the exact duration. Because of practical reasons we take T as the true duration, resulting in a time-aggregation bias in later parameter estimations. With respect to the time-aggregation problem, discrete event-history methods seem to be a natural mode to deal with this. But as Steinhage shows, these problems are not statistical ones but data problems, that cannot be solved by using discrete methods. The problem of time-aggregation is not only a problem of the dependent process but also of the parallel processes. In the working paper Steinhage shows the effects of time-aggregation and short spells on parameter estimation and discusses problems for empirical analyses using event-history methods.