Early warning signals for psychopathology

Many real-world phenomena share a common feature: the potential for sudden and difficult to reverse transitions into less desirable states. This applies to a wide range of domains, from ecosystems to social structures. The theory of dynamical systems has identified early warning signals (EWSs) that can precede these so-called critical transitions, enabling intervention before the system becomes locked into an undesired state. Recently, mental health researchers have explored the utility of EWSs for predicting transitions into unhealthy states (i.e. psychiatric disorders), which can be perceived as alternative stable states opposed to the healthy ones. In this work, we compare two viable approaches for early warning, a change point autoregressive model of order 1, CP-AR(1), and the kernel change point on running statistics, kcpRS. The comparison will be performed on a case study of a 57 years-old man with a history of major depression.