108 Chapter 5 with the lowest Extended Bayesian; Chen & Chen, 2008) is selected. The model with the optimal model fit is then visualized as a network using functionality from R package qgraph (Epskamp et al., 2012). For more detail on GVAR and regularization, see Epskamp et al. (2018) or Epskamp and Fried (2018). We present networks only on the contemporaneous timescale and not on a lag-1 timescale. This was done because 1) the lag-1 personality networks previously demonstrated within-person unreliability (Beck & Jackson, 2020, 2022) we deemed it unlikely that the processes we measured map onto one day-to-day timescale (e.g., Johns feeling restless on day 1 is more likely to make him blurt things out the same day than tomorrow) and 3) preliminary data-explorations confirmed the previous, showing long-term autocorrelations beyond lag-1 in various idiosyncratic ways (explicated in Section 3.2). 2.4.4 Between-person heterogeneity in within-person associations In the iterative process of estimating each of the idiographic networks, we generated 6x6 data matrices that contained the counts of all non-zero (i.e., significant) bivariate partial correlations at the individual level. These matrices provide the input for comparing within-person associations within the sample and within subgroups with the same personality profile. In each matrix the rows and columns reflected each of the six variables. Statistically significant positive partial correlations were saved in one diagonal of the matrix and the significant negative partial correlations in the other diagonal, so that heterogeneity in direction of partial correlations could also be assessed. One matrix contained the counts of all statistically significant associations between individuals of the whole sample, while the other matrices only contained counts of statistically significant associations for individuals within a specific subgroup. For example, one matrix reflects the counts of all sensation seekers. Functionality from qgraph (Epskamp et al., 2012) was used to visualize group-level networks that reflect the degree of homo- or heterogeneity within the sample and subgroups. 2.4.5 Within-person network variability and change In this last step we aimed to quantify and visualize within-person variability and change in the idiographic network structures over time. Using a sliding day window technique, we repeatedly estimated the idiographic network structure in segments of 30 consecutive days along the participant’s 60-day timeline. This means the network was first computed based on data-points between day 1 and day 30, then again between day 2 and day 31, and so on. For each participant, within each window, we calculated the node strength using package qgraph (Epskamp et al., 2012). Node strength quantifies how strongly a node is connected
RkJQdWJsaXNoZXIy MTk4NDMw