Improved estimation of autoregressive models through contextual impulses and robust modeling

The aim of the dynamic paradigm of affect research is to characterize daily life affective processes by means of intense longitudinal data and dynamic, typically autoregressive-type models. The contextual conditions accompanying and potentially influencing affective processes obviously form an important part of the picture. Especially distinct contextual events-ranging from salient daily events to major life events-are regularly assessed and their effects modeled. Such an explicit approach to studying context can however reach its limits, if relevant events are hard to define, measure or model. In that case one finds oneself in a situation where contextual events play out as hidden contaminators possibly obscuring the affective process of interest. Interestingly, specific forms of such contamination can also have beneficial effects in that they can trigger autoregressive dynamics and leverage their estimation. In this article, we take a closer look at this phenomenon focusing on events that exert short-lasting main effects on affective processes. We also demonstrate that robust autoregressive models have the capacity to outperfom classical models given this and other forms of contextual contamination as they not only capitalize on positive leverage effects but also mitigate the negative obscuring effects contextual events might have. (PsycInfo Database Record (c) 2025 APA, all rights reserved).