Figure 3

Building the psychological landscape, from raw data to the determination and the causal analysis of its dynamical patterns. A. The whole dataset is constituted of various data layers collected on persons, which include behavioral, biological, genetic, and imaging data. It may exhibit complex relationships that cannot be preserved or explored appropriately with classic statistical tools. B. Reduced dataset produced by nonlinear reduction techniques (NLDR), which allow minimal complexity loss (step 1). C. A raw map is produced by pattern recognition techniques and robust clustering (step 2). Homogeneous groups of persons may be identified, according to similar behaviors and psychological traits. Such groups are termed nosological areas. D. Zooming in on a specific nosological area. E. Applying algorithms for attractor reconstruction should allow the dynamic pattern of this area to be identified. F. The causal pathways leading to an attractor can be analysed with Bayesian networks or similar methods (steps 4 and 5). The links of the dynamic pattern to other items from the whole dataset, e.g., age and sex, can be also analysed. G. Once the main nosological areas are identified, pathways between these areas can be searched for. Bayesian networks can be used as well.