Skip to main content

Table 2 Differences in bicycle commuting (%) in 2010 vs 2000a by trail access and potential useb

From: Municipal investment in off-road trails and changes in bicycle commuting in Minneapolis, Minnesota over 10 years: a longitudinal repeated cross-sectional study

 

Percentile

Unadjusted

Multivariable-adjusted, model 1c

Multivariable-adjusted, model 2d

Distance from trail system (km)

25th (1.08 km)

3.00 (2.28, 3.72)

2.54 (0.71, 4.38)

2.03 (0.13, 3.93)

50th (2.83 km)

2.50 (1.94, 3.06)

1.99 (0.20, 3.77)

1.88 (0.10, 3.66)

75th (5.91 km)

1.62 (0.95, 2.29)

1.01 (-0.90.00, 2.92)

1.62 (-0.41, 3.65)

p-value, trend

<0.01

<0.01

0.63

Proportion of commuting trips crossing the trail system

25th (0.11)

1.20 (0.6, 1.95)

0.73 (-1.14, 2.61)

0.95 (-1.01, 2.90)

50th (0.29)

2.37 (1.84, 2.90)

1.83 (0.07, 3.59)

1.89 (0.08, 3.69)

75th (0.42)

3.22 (2.52, 3.92)

2.63 (0.79, 4.46)

2.57 (0.53, 4.60)

p-value, trend

<0.01

<0.01

0.06

  1. aDifferences were obtained using the ‘margins’ post-estimation command following repeated-measures random effects linear regression models (-xtreg-) in Stata
  2. bLevels of predictor variables (a. distance from the trail system and b. proportion of work-related trips that cross the trail system.) reflect the 25th, 50th, and 75th percentiles of the variable distribution for combined 2000 and 2010 data
  3. cRegression models included either distance to trail system (analysis a) or proportion to work-related trips that cross the trail system (analysis b), and adjusted for time-varying and tract-level covariates: total work-related trips, intersection density, population density, median household income, professional workforce, workforce aged 18-34 years, total length of bicycle lanes, maximum reach of bicycle lane network, maximum reach of network comprising both bicycle lanes and off-road trails, and the time-invariant variable for commuting by bicycle in 1990. Estimated effects for changes for all time-varying variables were modeled by including a year*variable interaction term
  4. dRegression models included both distance to trail system and proportion of work-related trips that cross the trail system, controlling for the same set of covariates as in 3