Caregivers experienced substantial psychological strain during the first year of the COVID-19 pandemic, while religious and spiritual resources were frequently invoked as possible supports for coping. The present study asks a question that is different from directional mediation: when all repeatedly measured constructs are considered simultaneously, which associations remain conditionally connected across the six-month observation period? The analytic input was a published nine-variable correlation matrix for 549 caregivers assessed in May, September, and November 2020. The matrix included initial pandemic disruption, religion/spirituality at the first and third assessments, positive coping at all three assessments, and psychological distress at all three assessments. An extended Bayesian information criterion graphical lasso was used to estimate a sparse Gaussian graphical model. The primary penalty setting retained 22 of 36 possible edges; 13 had an absolute regularized partial correlation of at least 0.05. The conditional structure was dominated by temporal continuity within religion/spirituality (\(\rho=0.901\)), coping (\(\rho=0.326\)–\(0.455\)), and distress (\(\rho=0.203\)–\(0.470\)). Distress and coping remained inversely connected within every assessment (\(\rho=-0.106\) to \(-0.155\)). In contrast, marginal religion/spirituality–coping associations attenuated to \(|\rho|\leq0.022\) after simultaneous conditioning. Initial disruption retained conditional links with concurrent distress (\(\rho=0.301\)) and November distress (\(\rho=0.138\)). The matrix remained positive definite, and all selected edges persisted across 200 perturbations bounded by the reported three-decimal rounding intervals. These findings indicate that temporal stability and the contemporaneous distress–coping relation constitute the principal conditional architecture of the published correlations. They do not establish causal direction and do not invalidate mediation estimates obtained from participant-level longitudinal models. Instead, they identify which relations are unique at the aggregate correlation level and which are largely shared with repeated measurements of the same constructs.
Caregivers experienced substantial psychological strain during the first year of the COVID-19 pandemic, while religious and spiritual resources were frequently invoked as possible supports for coping. The present study asks a question that is different from directional mediation: when all repeatedly measured constructs are considered simultaneously, which associations remain conditionally connected across the six-month observation period? The analytic input was a published nine-variable correlation matrix for 549 caregivers assessed in May, September, and November 2020. The matrix included initial pandemic disruption, religion/spirituality at the first and third assessments, positive coping at all three assessments, and psychological distress at all three assessments. An extended Bayesian information criterion graphical lasso was used to estimate a sparse Gaussian graphical model. The primary penalty setting retained 22 of 36 possible edges; 13 had an absolute regularized partial correlation of at least 0.05. The conditional structure was dominated by temporal continuity within religion/spirituality (\(\rho=0.901\)), coping (\(\rho=0.326\)–\(0.455\)), and distress (\(\rho=0.203\)–\(0.470\)). Distress and coping remained inversely connected within every assessment (\(\rho=-0.106\) to \(-0.155\)). In contrast, marginal religion/spirituality–coping associations attenuated to \(|\rho|\leq0.022\) after simultaneous conditioning. Initial disruption retained conditional links with concurrent distress (\(\rho=0.301\)) and November distress (\(\rho=0.138\)). The matrix remained positive definite, and all selected edges persisted across 200 perturbations bounded by the reported three-decimal rounding intervals. These findings indicate that temporal stability and the contemporaneous distress–coping relation constitute the principal conditional architecture of the published correlations. They do not establish causal direction and do not invalidate mediation estimates obtained from participant-level longitudinal models. Instead, they identify which relations are unique at the aggregate correlation level and which are largely shared with repeated measurements of the same constructs.
