Not straightforward: Mediation and networks in continuous time
The advent of smartphone technology has led to a huge increase in the availability of intensive longitudinal (also known as time series, ambulatory assessment, experience sampling methodology or ESM) data. In psychology in particular this type of data is increasingly being used to model psychological processes or disorders as dynamic systems. There a wide range of models used to do this, but the most popular models are based on the analysis of lagged relationships between variables measured at different occasions. Such approaches form the core of both longitudinal mediation analysis and dynamical network modeling.
However, it is well-known that these relationships depend on the amount of time that elapses between measurements, such that, among other problems, results cannot be generalized to other lags. An innovative and elegant solution to this problem is to adopt a continuous time (CT) modelling approach, based on the use of differential equation models. This shift to a CT modelling approach however also entails a shift in the perspective with which the dynamic systems we are interested in are viewed. This leads to many major implications regarding the calculation and interpretation of for instance path-specific effects, and other causal or quasi-causal notions. The current project is concerned with developing a CT approach to dynamical network analysis, and tackling the most urgent problems that arise when applying the CT perspective to mediation and network analysis.
Dr Ellen Hamaker & Prof. Peter Van der Heijden
NWO Research Talent Grant
1 September 2015 – 31 August 2019