Tatjana StadnitskiMultivariate Time Series Analyses for Psychological Research

About this book deutschenglish

Mental functions and behaviors are often connected to dynamic systems. A dynamic system consists of two or more interrelated processes. This paper introduces multivariate procedures such as VAR (Vector Autoregressive), SVAR (Structural Vector Autoregressive), VEC (Vector Error Correction), and SVEC (Structural Vector Error Correction) models to psychologists and demonstrates their application with examples employing the open source statistical software JMulTi and R. The described procedures treat time series under study as a system revealing interdependencies in their dynamics and allowing testing of causal hypotheses. In contrast to the majority of methodological papers from this field, explanations and illustrations considered here are not restricted to stationary models. Many psychological processes are integrated (i.e., extremely unstable). In such nonstationary cases, it is difficult to discern whether time series under study are interconnected or independent because of spurious regression. The Johansen procedure described in the present paper provides a reliable diagnostic tool to distinguish between different types of dynamic relationships: stationary systems, cointegrated systems, and independent integrated variables. Elaborate empirical demonstrations and detailed documentation of the conducted analyses aim to enable practitioners to implement the discussed multivariate strategies in applied settings.