
Project
SEM 2.0: Towards personalized multi-disciplinary treatment plans
The next-generation approach to research is based on intensive data collection and complex models. Due to the digital revolution, measurement tools have become technologically advanced (e.g., wearable devices that track one’s physiology and location, genetic sequencing tools, or digital assessment tools used in educational testing) and available data has increased. This often results in data that has more variables than cases, commonly referred to as high-dimensional low-sample-size (HDLSS) data. On the other hand, research paradigms have shifted towards multidisciplinary approaches. As a result, behavior and cognition are now examined not only through a psychological perspective but also through various other disciplinary perspectives, including environmental, social, clinical, and biomolecular perspectives (see, e.g., van de Poll-Franse et al., 2022). This leads to the understanding that a person-centred approach is necessary, as behavior represents a complex interplay of multiple disciplinary domains. For instance, clinical and medical psychologists realize that there are large inter-individual differences in the treatment plans needed: for some patients poor outcomes mainly depend on behaviour (e.g., unhealthy lifestyle), for some on genetic constitution (e.g., immune deficiencies), and for some on a subtle interplay between genetic susceptibility and triggering/protective behaviour. Hence, researchers need more accurate explanatory models that account for individual differences.
Structural equation models (SEM, Kline, 2015) are particularly powerful tools for building explanatory models of behaviour and cognition since they model both measurement model and structural model. However, currently available structural equation methods do not meet the current research needs that come with large-scale multidomain data and person-centred approaches. First, these methods do not scale up to the HDLSS setting. The parameter estimates are known to be unreliable when sample sizes are small and solutions may even not compute (Rosseel, 2020). Second, they do not account for the multidomain structure of the data. To obtain a good understanding of how behaviour and cognition depend on the multiple domains, domain-specific mechanisms should be disentangled from joint mechanisms. This implies constructing latent variables that account for the grouped or block structure of the variables (with each group/block referring to a particular domain). Third, currently available SEM methods do not support personalized approaches as this implies modelling (almost) at the individual level, which results in an extreme case of the HDLSS setting. And last, the latent variable or factor scores are indeterminate in the structural equation model whereby the method has very limited value for predictive modelling (Hair & al., 2017). Despite, latent-variable based predictive modelling has been advocated as the way forward in dealing with measurement error to leverage the potential of machine learning based prediction (Jacobucci & Grimm, 2020; McNeish & Wolf, 2020).
This project develops a regularized exploratory structural equation modelling method, regularized ESEM, that addresses the challenges posed by the modern research paradigms based on intensive collections of multidisciplinary data and a desire to account for substantial heterogeneity in the underlying mechanisms which also accounts for the need of out-of-sample prediction. The method
1. scales up with complexity such that it can be used when sample sizes are small, when the number of variables is high, and when differences among many (unknown) groups exist.
2. yields insightful results by imposing simple structure not only within domain-specific data but also across domains to reveal the joint and domain-specific mechanisms that shape behaviour and cognition.
3. takes a major leap towards person-centred modelling by building on regularization to avoid overfitting and on convex optimization and distributed computing to address computational efficiency.
4. focuses both on explanation and prediction.
Supervisors
Prof. Dr. Katrijn van Deun
Dr. Zsuzsa Bakk
Financed by
NWO Talent Programme
Period
1 October 2024 – 1 October 2028