A cross-sectional study of demographic, environmental and parental barriers to active school travel among children in the United States

Table 3 Estimates for assessing associations between active school travel and each of covariate in the base and final models

	Base model			Final model
	Estimate	t-value	p	Estimate	t-value	p
Intercept	−0.0989	−0.17	0.8686	−0.4832	−2.13	0.0353
Distance	0.0275	6.00	< 0.0001	0.0301	16.60	<0.0001
White	Not in b/c bivariate p > 0.15
African American	−0.0006	−0.15	0.8795
Hispanic	0.0052	2.58	0.0110	0.0059	4.23	<0.0001
Temperature (°F)	0.0030	1.17	0.2429	0.0034	1.53	0.1285
Walk/bike-ability	0.0024	1.38	0.1695	0.0031	2.26	0.0255
Safety and weather ^*	−0.0160	−1.53	0.1280	−0.0245	−4.27	<0.0001
Suitability of the route to school	0.0102	1.07	0.2868	0.0182	3.19	0.0018
No walking companion	Not in b/c covary with Safety and Weather (r = 0.821)
Children’s resistance	−0.0453	−4.22	<0.0001	−0.0438	−4.83	<0.0001
Parents’ perceived active school travel norms	−0.0005	−0.09	0.9292
Children’s perceived barriers	−0.0096	−0.85	0.3975
Children’s perceived active school travel norms	0.0018	0.36	0.7184

All models were adjusted for clustering (school*time) and distance (% within 1 mile of school).
Base model was reduced in a stepwise fashion in building final model. One variable was removed at each step based on the significance of the parameter estimates (largest p-value greater than 0.15) and the change in the log-likelihood (smallest change when variable removed). The stepwise reduction continued until p-values for all parameters were < 0.15 and/or change in log-likelihood > 2.00 (df = 1) for any variable removed. This process resulted in removing 4 variables: percentage of African American students, children’s perceived barriers, parents’ perceived active school travel norms, and children’s active school travel norms.
*The variable “safety and weather” was strongly correlated with “no walking companion” (r = 0.821); for this reason only one was retained in the base model.

ISSN: 1479-5868