Online Shopping Addiction in Association with Depression, Anxiety, and Stress: Exploring Independent and Relative...

Psychological distress is associated with problematic online shopping, but the relative and independent associations of depression, anxiety, and stress require careful evaluation without causal interpretation. This study examined their associations with continuous Online Shopping Addiction Scale (OSAS) scores after accounting for sociodemographic characteristics and online shopping frequency. A cross-sectional convenience sample included 438 adults in Türkiye (54.0% female among respondents reporting gender; mean age = 33.79, SD = 10.24); a total of 317 participated online and 121 completed printed questionnaires face to face. Pearson correlations were estimated in the full sample. Hierarchical regressions used a common complete-case sample (n = 375). Because residual diagnostics indicated heteroscedasticity and tail departures, HC3 robust standard errors and 95% confidence intervals were reported. Anxiety, depression, and stress showed statistically significant but modest correlations with OSAS total scores (r = 0.34, 0.32, and 0.27, respectively; all p < 0.001). In separate adjusted models, anxiety (ΔR² = 0.070; β = 0.283), depression (ΔR² = 0.068; β = 0.276), and stress (ΔR² = 0.042; β = 0.217) each increased explained variance (all p < 0.001). In the combined model, the DASS-21 block added ΔR² = 0.091; anxiety (β = 0.206, p = 0.002) and depression (β = 0.200, p = 0.001) had comparable independent associations, whereas stress did not (β = -0.050, p = 0.473). Online shopping frequency had the largest standardized coefficient among concurrent covariates (β = 0.280), but it is conceptually proximal to the OSAS outcome. The combined model explained 31.6% of the variance (adjusted R² = 0.291). Recruitment-mode comparisons were small and nonsignificant after Holm correction. Depression and anxiety were independently associated with higher OSAS scores, while most outcome variance remained unexplained. The findings do not establish temporal prediction, causation, or clinical diagnosis and should be interpreted in light of self-report measurement, construct overlap, residual confounding, and the digitally engaged convenience sample.