Mihai Constantin

University of Tilburg
Department of Methodology and Statistics


Tools for Aiding Empirical Research Based on Intensive Longitudinal Data

In the recent years, two increasingly popular topics have occupied the front lines of
psychopathological research. First, is the conceptualization of mental disorders as complex systems in which symptoms interact to produce patterns that lead to the emerge of mental disorders. This view is known as the network approach, or the network theory of mental disorders (Borsboom, 2017; Cramer et al., 2016; Cramer, Waldorp, van der Maas, & Borsboom, 2010). Another active direction in clinical research is the use of intensive longitudinal data (???). These data—as opposed to, for example, crosssectional or panel data—are regarded as gateway to designing personalized interventions and are usually acquired via the experience sampling methodology (ESM; Larson & Csikszentmihalyi, 2014) in which participants are asked to answer relatively simple items (e.g., “I feel cheerful”) multiple times per day, at random moments, over the course of several days (Trull & Ebner-Priemer, 2009). Numerous methods
exist for modeling ESM data (Hamaker, Asparouhov, Brose, Schmiedek, & Muthén, 2018; Hamaker, Ceulemans, Grasman, & Tuerlinckx, 2015), however, a popular choice in social sciences is the first order vector autoregressive model, in short VAR (1) (Brandt & Williams, 2007). Within a VAR (1), each variable is regressed on itself and all other variables at the previous time point (i.e., also called lag 1; see Figure 1a). Recently, network psychometricians turned their attention to such methods in order to construct intra-individual networks (Bringmann, Lemmens, Huibers, Borsboom, & Tuerlinckx, 2015; Bringmann et al., 2013; Epskamp et al., 2018), in which the links between the measured variables denote temporal predictive relations (e.g., experiencing bodily discomfort at a time point predicts being nervous at the next time point; see Figure 1).
The aim of this project is to further advance the use of ??? to facilitate personalized, networkbased research. We take a pragmatic stance and tackle issues that are immediately relevant to researchers who want to design studies based on ??? data. More specifically, we focus on (1) the power requirements for sensibly using such methods, (2) evaluating and developing effect sizes that suit different research question that employ network models and ??? (3) the issue of measurement error when using noisy ??? data, and (4) the many degrees of freedom with respect to the choices researchers make when collecting and analyzing such data.


Prof. dr. J. Vermunt. dr. A.O.J. Cramer, dr. N.K. Schuurman

Financed by

University of Tilburg


1 September 2018 – in progress