Maarten Marsman, assistant professor in psychological methodology at the University of Amsterdam, says that a network approach to explain our cognitive development and mental problems is becoming increasingly common in the social sciences. ‘For a long time, for example, it was thought that our personality and intelligence – as well as mental problems such as depression – were found somewhere in our brain. This would mean that differences between you and me were explained by the fact that my brain deviated from yours. This way of thinking is now increasingly giving way to the revolutionary theory that cognition and mental problems arise as a result of observable factors or symptoms that interact with one another. For example, people who do not sleep well cannot concentrate and may get irritated at work. In turn, this will affect your mood and how well you sleep.’
Marsman sees this focus on the interaction between factors in complex systems reflected in the scientific arena as a whole. ‘More and more attention is being paid to this factor in all kinds of social issues, and less and less attention is being paid to finding a single identifiable explanation that probably doesn’t even exist. Although a network approach can reveal patterns and relationships between variables, statistical innovations can barely keep pace with the growth in network studies. As a result, data can be misinterpreted.’ Marsman plans to offer a solution to this.
A new framework for statistical analysis
Marsman received an ERC Starting Grant to develop a new framework for the analysis of psychological and other networks. ‘How can I ensure that I have the right model for the specific data at my disposal? And how do I test whether or not there is a relationship between specific nodes in the network, where I can intervene? Between poor sleep and concentration problems in the event of depression, for example.’ Marsman will also develop methods to investigate the impact of interventions on the network. ‘Because it is still very difficult to find out which way network relations go, and what influences what.’
In order to make the predictions as robust as possible, Marsman applies a so-called Bayesian approach that also incorporates doubt and uncertainty about certain relationships. ‘All possible network structures, which combine different factors to explain a particular issue, will initially have the same opportunity to prove their plausibility. I then weigh which models lead to more and less plausible results and take this weighting forward. A less plausible model is therefore not ruled out, but instead contributes by revealing connections that seem less strong and more uncertain.’
Making statistics available for free in JASP
Marsman will share his methodology with the world via JASP. JASP is an open-source platform, developed by colleague Eric-Jan Wagenmakers, that makes methodology ‘easy, insightful, and user friendly’ for researchers, free of charge. ‘I will also test my methodology outside my own field to ensure that it incorporates the experiences of different users of the network approach.’ Marsman is looking forward to the moment when he can include an initial implementation in JASP. ‘People currently have too much confidence in the methods available, which means that data can be misinterpreted. Not only will we soon be solving an isolated problem, we will also be establishing a methodology that solves a great many problems in depth. I’m very proud of that.’
In the coming period, Marsman will be putting together his team of two PhD candidates: a postdoc and a programmer. Interested? Keep an eye on the UvA’s vacancy page.