A Component-Level Network Analysis of Emotional Symptoms, Sleep Problems, and Physical Activity Identity at a Chinese University

This cross-sectional study estimated a component-level network of depression, anxiety, and stress dimensions, sleep components, and physical activity identity among students at one Chinese university. Using convenience and snowball sampling, 3762 students at Zhejiang University completed the DASS-21, PSQI, and MIPAI-25. Regularized partial-correlation networks were estimated with EBICglasso (gamma = 0.50) using automatic correlation selection. Strength, bridge strength, signed bridge expected influence, and predictability were calculated. Network accuracy and stability were examined with 1000 bootstrap resamples, and sex differences were assessed using a Network Comparison Test with 1000 permutations. Scores above the prespecified screening cutoffs were observed for depression (76.0%), anxiety (79.5%), stress (62.8%), and poor sleep quality (80.2%). In the DASS-PSQI network, hypnotic medication use had the largest overall strength point estimate (1.126), whereas depression and anxiety subscales had the largest bridge-strength point estimates (0.121 and 0.093). In the exploratory three-domain network, hypnotic medication use again had the largest strength estimate (1.131); MIPAI acknowledgment had the highest bridge-strength point estimate (0.134), closely followed by depression (0.130), and remained the highest-ranked MIPAI dimension in the mixed graphical model. Its individual cross-domain edges were small, and their specific pattern was model-dependent. Overall network structure did not differ significantly by sex (M = 0.087, p = 0.117), whereas global strength was higher in males than females (7.11 vs. 6.83; p = 0.012); no individual edge remained significant after Holm correction. These findings describe cross-sectional conditional associations and do not establish causal or protective effects.