Issue 2
Vol. 43
2001

Identifying Boolean Networks in Multivariate Time Series:
The prediction of contact behavior in friendship dyads
Torsten Reimer, Ann Elisabeth Auhagen

Summary
Boolean Network Time Series (BNTS) provides a method of analysing multivariate time series on the basis of Boolean functions. In contrast to other time series analyses, BNTS determines a Boolean network, i.e. any variable serves as a criterion and as a potential predictor simultaneously. For each variable of a vector such a combination of predictors and Boolean functions is ascertained that leads to the best prediction of the same vector with respect to the specified lags assigned to the criterion and the predictors. In this chapter the general procedure determining a Boolean network is outlined by using a study of contact behavior in friendship as an example. The analysis contains a variation of numbers of predictors and time lags and shows how to aggregate over several Boolean networks. Loglinear models are used to test the fit of a specified Boolean network and to test hypotheses on an aggregate level.

Dr. Torsten Reimer
Department of Psychology
Free University of Potsdam
Germany

Dr. Ann Elisabeth Auhagen
Department of Psychology
Free University of Berlin
Habelschwerdter Allee 45
D-14195 Berlin
Germany