Table 2.
Associations of age, educational attainment and sex with physical activity (PA) outcomes: Main effects for entire sample
| PA outcome | n | Age | Educational attainment (reference: < high school graduate) |
Sex (reference: male) |
|
|---|---|---|---|---|---|
| High school graduate or some college | College degree or higher | Female | |||
| eb (e95% CI) | eb (e95% CI) | eb (e95% CI) | eb (e95% CI) | ||
| Transport-related PA (no significant curvilinear associations) | |||||
| Yes, ≥10 min/week a | 13,233 | F(2, 13.208) = 7.42*** | |||
| linear: | 0.991 (0.987, 0.994)*** | 0.95 (0.83, 1.09) | 1.15 (1.00, 1.33) | 1.02 (0.94, 1.11) | |
| Non-zero min/week b | 10,166 | F(2, 10,141) = 9.24*** | |||
| linear: | 1.003 (1.001, 1.004)** | 1.00 (0.94, 1.06) | 0.89 (0.84, 0.95)** | 0.87 (0.84, 0.91)*** | |
| Yes, ≥150 min/week a | 13,233 | F(2, 13,208) = 0.23 | |||
| linear: | 1.001 (0.998, 1.004) | 1.04 (0.93, 1.17) | 1.02 (0.91, 1.16) | 0.90 (0.84, 0.98)** | |
| Leisure-time PA | |||||
| Yes, ≥10 min/week a | 13,233 | F(2, 13,209) = 146.32*** | |||
| linear: | 0.750 (0.549, 1.025) | 1.51 (1.35, 1.70)*** | 2.16 (1.90, 2.45)*** | 0.90 (0.83, 0.97)** | |
| curvilinear (smooth term) c: |
χ2(3.94) = 46.96*** (see Fig. 1a) |
– | – | – | |
| Non-zero min/week b | 9427 | F(2, 9403) = 3.84* | |||
| linear: | 1.025 (0.854, 1.231) | 1.02 (0.96, 1.09) | 0.95 (0.89, 1.02) | 0.87 (0.84, 0.91)*** | |
| curvilinear (smooth term) c: |
F(3.68, 9399.30) = 5.42*** (see Fig. 1b) |
– | – | – | |
| Yes, ≥150 min/week a | 13,233 | F(2, 13,209) = 81.11*** | |||
| linear: | 0.889 (0.697, 1.134) | 1.43 (1.27, 1.61)*** | 1.75 (1.55, 1.98)*** | 0.87 (0.81, 0.94)*** | |
| curvilinear (smooth term) c: |
χ2(3.53) = 18.56*** (see Fig. 2) |
– | – | – | |
95% CI = 95% confidence interval; exp.(b) exponential function of regression coefficient; exp.(95% CI) = exponential function of confidence interval; n = sample size
a Generalized additive mixed model (GAMM) with binomial variance and logit link functions, for which exp.(b) is interpreted as odds ratio
b GAMM with Gamma variance and logarithmic link functions, for which exp.(b) is interpreted as the proportional increase in PA associated with a 1 unit increase in the predictor
c The approximate statistical significance of the smooth term is based on chi-squared (χ2) and F-ratio statistic for GAMMs with binomial and Gamma variance function, respectively. All models adjusted for city, working status, marital status, administrative-unit-level socio-economic status and walkability
* p-value ranging from ≤.05 to >.01; ** p-value ranging from ≤01 to >.001; *** p ≤ .001