122 Chapter 5 Our primary recommendation for future research is to include situational information in the data collection. With information about the context it is possible to model personality as consisting of multi-level networks, where quick-varying situational features and behaviors, intermediately varying moods and/or evaluations and slow-changing internal processes are placed in separate but interacting layers of the model (Kivelä et al., 2014; cf. de Boer et al., 2021). This would allow the multiple relevant timescales to be studied at once. Importantly, the solution to non-stationarity is not just statistical but also theoretical. Estimating a stable idiographic personality network may never be achieved when the timeframe, timescale(s) and situational features (Cervone, 2005; Mischel & Shoda, 1995) remain unspecified. It is our contention that even the most advanced statistical models will not solve the problems demonstrated in this study. Therefore the ultimate challenge for personality researchers who wish to employ idiographic networks to study individual differences is to start with further theory building. The current study has some limitations. Particularly around idiographic network variability over time we build on limited evidence, which warrants cautious conclusions. That is, we cannot automatically assume our nonstationarity findings are generalizable, because we do not know how representative our sample is for the population in terms of personality network stability. Current personality theory, on the other hand, also does not inform us about which demographic variables influence personality stability. Future research should explore this further in different samples with different demographics. Second, we based our networks on 60 timepoints per individual. A recent simulation study suggest that this may be underpowered, which makes for an increased possibility we mistake noise for true between- or within-person heterogeneity (Hoekstra et al., 2022). Future research is encouraged to compare idiographic networks estimated from sufficiently large segments (e.g., assuming it is feasible, windows with 300 data-points within a timeseries with 600 time-points). Importantly, when the data-generating processes are in fact non-stationary, variability over time remains a realistic prospect and it will still be unclear what inferences we can derive from the average network. In a more general sense, the current paper illustrates that conclusions about personality based on between-person structures (such as the attributing of SURPS profiles) and within-person structures (idiographic network models) are not compatible (cf. Brose et al., 2015). This stresses the need for an idiographic perspective complementary to the nomothetic perspective that dominates research on personality. Idiographic networks are one intuitive
RkJQdWJsaXNoZXIy MTk4NDMw