Disentangling Between- and Within-Person Effects of Nonphysical Aggression and Nonphysical Victimization Over Time in Adolescence.

Self-reported knee pain is one of the most common musculoskeletal complaints among community-dwelling older adults and a leading contributor to disability, reduced physical activity, and impaired quality of life [,]. The Global Burden of Disease 2021 study estimated that 595 million people worldwide were living with osteoarthritis in 2020, corresponding to 7.6% of the global population, with knee being the most common site; compared with 2020, the number of cases of knee osteoarthritis is projected to increase by 74.9% by 2050, alongside continued increases in population aging and obesity prevalence [,]. Identifying which factors are associated with self-reported knee pain in community settings, before structural disease and clinical referral, is therefore a continuing research priority. Knee pain in older adults is conceptualized as multidimensional. The () framework, built on the biopsychosocial model, organizes health into interacting domains spanning body functions and structures, activities and participation, and environmental and personal factors [-]. In knee osteoarthritis populations, physical performance, psychological status, sleep, and environmental context each carry independent associations with pain experience and functional limitation [-]. The specifies that health outcomes such as knee pain arise from a structure in which some variables show direct associations with the outcome and others act through intermediate variables. Although this direct-versus-indirect distinction is part of the framework, prior studies of knee pain have not quantitatively separated the two types of association in a single analysis. As a result, it remains unclear which domains carry direct associations with knee pain in older adults and which contribute through other variables in the multidimensional structure. Two analytical paradigms have been applied to multidimensional pain data, but they answer different questions and have typically been used in isolation. Machine learning (ML) classification with Shapley Additive Explanations (SHAP)–based explanation quantifies the marginal predictive importance of each feature, capturing potentially nonlinear contributions that traditional regression may miss [,]. Recent reviews of ML applications in knee osteoarthritis highlight the rapid growth of this methodology across clinical, structural, and surgical endpoints, while also noting that most studies focus on imaging-based outcomes and that few have addressed self-reported pain or nonimaging multidimensional determinants in older adults [,]. Partial-correlation network analysis characterizes the conditional dependency structure among variables, with centrality indices summarizing how each variable connects to the broader system after controlling for all others [,]. When the two paradigms are reported jointly on the same data, it becomes possible to determine whether a variable’s contribution to the model output is mirrored by a direct conditional dependency with the outcome, or whether that contribution is accounted for by other variables in the network. This joint use operationalizes the direct and indirect pathways described by the , but applications of this combined analysis to knee pain remain limited [,]. Mobile health measures of physical performance, such as smartphone-based gait and sit-to-stand assessments [], need to be integrated with questionnaire-based body function and environmental variables within a single -aligned analysis.