Skip to main content

Table 4 Results from multilevel ordinal logistic regression models of differences in route lengths, and percentage overlap between GIS and GPS routes

From: How well do modelled routes to school record the environments children are exposed to?: a cross-sectional comparison of GIS-modelled and GPS-measured routes to school

 

Adjustment for straight-line distance only

 

Differences in Length

% of GPS route within 50 m GIS route buffer

Best fit

Univariate associations

Best fit

   

95% CI

  

95% CI

  

95% CI

  

95% CI

 
  

OR

Lower

Upper

 

OR

Lower

Upper

 

OR

Lower

Upper

 

OR

Lower

Upper

 

Straight-line distance

 

2.25

2.02

2.52

**

1.37

1.16

1.62

**

1.11

1.01

1.20

*

0.82

0.72

0.94

**

Journey type (ref to sch)

From school

1.14

0.85

1.54

     

0.76

0.56

1.02

     

Sex (ref male)

Female

0.41

0.23

0.74

*

    

3.44

2.11

5.62

**

    

Travel mode (ref car)

Bus

11.32

4.96

25.86

**

8.32

3.08

22.50

**

1.02

0.54

1.92

 

1.93

1.00

3.72

 

Bike

2.40

0.88

6.54

 

0.55

0.17

1.80

 

1.33

0.45

3.90

 

2.79

1.07

7.31

*

Foot

0.06

0.03

0.13

**

0.15

0.06

0.36

**

0.54

0.29

1.01

 

1.42

0.63

3.19

 

Home location (ref urban)

Town & Fringe

1.54

0.80

2.96

 

0.75

0.34

1.64

 

0.42

0.25

0.72

**

0.19

0.11

0.35

**

Village etc.

3.49

1.87

6.49

**

4.62

1.74

12.22

**

0.77

0.47

1.29

 

0.26

0.10

0.74

*

School location relative to school (ref no diff)

School less urban

0.49

0.15

1.63

     

0.08

0.04

0.17

**

0.98

0.35

2.79

 

School more urban

2.27

1.30

3.96

**

    

3.26

2.03

5.26

**

3.98

1.49

10.63

**

Variance partition coefficients

    

Variance

VPCa

     

Variance

VPCa

 

 Level 1 (Route)

     

3.29

32.10%

     

3.29

16.72%

 

 Level 2 (Participant)

     

6.24

60.91%

     

12.41

63.09%

 

 Level 3 (School)

     

0.72

7.00%

     

3.97

20.19%

 
  1. The distributions of both outcome variables did not allow for linear modelling, so both were banded into tertiles. Outcomes are presented as odds ratios for moving up a tertile category. Difference in Length Tertiles; GIS longer, GPS up to 600 m longer, GPS >600 m longer.% overlap tertiles: 0.1%-33%, >33%- < 73%, ≥73%-100%. *p < 0.05, **p < 0.01, aVariance Partition Coefficient.