Table 3.

Multinomial multi-level logistic regression model building for perceived stress (n = 2537).

Level	Variables	Model 1 −2LL (full –reduced) (p value)	Model 2 −2LL (full –reduced) (p value)	Model 3 −2LL (full –reduced) (p value)	Model 4 −2LL (full –reduced) (p value)	Model 5 −2LL (full –reduced) (p value)	Model 6 −2LL (full –reduced) (p value)
Level 2	City	43.7 (<0.0001)	11.7 (0.067)	13.03 (0.042)	12.03 (0.061)	11.5 (0.072)	10.04 (0.123)
	City  + school type (M/F)		52.7 (<0.0001)	46.2 (<0.0001)	50.2 (<0.0001)	50.08 (<0.0001)	59.9 (<0.0001)
	City  + school type (M/F)  + school level			23.7 (0.008)	25.5 (0.004)	22.2 (0.014)	22.7 (0.012)
Level 1	City  + school type (M/F)  + school level  + GAS				54.3 (<0.0001)	50.4 (<0.0001)	43.4 (<0.0001)
	City  + school type (M/F)  + school level  + GAS  + age  + academic grade					3.2 (0.195) 61.4 (<0.0001)	3.2 (0.194) 43.4 (<0.0001)
	City  + school type (M/F)  + school level  + GAS  + age  + academic grade  + screen time  + smoking  + physical activity  + diet						10.3 (0.112) 13.4 (0.001) 17.1 (<0.0001) 16.7 (<0.0001)

LL: log-likelihood ratio.