162 Chapter 7 not possible, analogous to how the idiographic networks could not summarize multiple non-stationary processes (Chapter 5). The same is true for the second participant (middle panel Figure 1) with a time-varying variance in fear. Of all three participants, the score of the third participant (bottom panel) most accurately conveys the timeframe, as the process was most stationary. Thus, the impossibility of summarizing non-stationary processes casts doubt on measurements in retrospective surveys—the same measurements that serve as input for analyses upon which conclusions are based. Importantly, the crux is that in real life, neither the participant nor the researcher will have a clue about the processes' (non)stationarity upon which participants base their responses. The impossibility of summarizing a nonstationary process has nothing to do with retrospective bias, nor can it be passed off as measurement error, nor can it be attributed to the psychometric qualities of the survey. It is simply impossible. Yet, when participants are asked a question, they will answer. This also forms a limitation of daily monitoring (Chapter 4): if the morning was very fearful and the afternoon fearless, then there is no summary that captures the whole day. Nevertheless, EMA data can at least show (non)-stationarity across days or hours, but in pre-posttest designs this remains completely unknown. Potentially invalid summaries of non-stationary idiographic processes are inputted in datasets, aggregated within participants to create a sum-score of multiple variables (e.g., substance use severity in Chapter 2 or the YSR domain rule-breaking problems in Chapter 3). These aggregated domain scores are subsequently aggregated between people to test average effects. Abstractions upon abstractions detach from any concrete behavior in any particular context. I consider it very problematic that it is unclear how well such abstractions reflect any participant at any point in time. Complement EMA with other data sources The most pressing recommendation for the study of change in people with a mild intellectual disability is to employ EMA research designs instead of longitudinal designs with a few measurements. This method makes it easier for participants to provide unbiased accounts of their daily lives, paint the change process in far more detail than pre-posttest designs and may guide analytical decisions on handling (non)stationarity. Chapter 4 illustrated that feasibility and acceptability among people with a mild intellectual disability diagnosis is good, especially when the participant has access to their own mobile phone to answer the diaries. Method feasibility was recently echoed by research (Bakkum et al., 2024) in which participants with a mild intellectual disability tested three different EMA apps:
RkJQdWJsaXNoZXIy MTk4NDMw