Date controls gave smaller gaps in casual trip share between weather classes for Capital Bikeshare hours from 06:00 to 22:00 in 2011–2012
We tested 12,375 Capital Bikeshare hours from 06:00 to 22:00 in 2011–2012. We measured the casual share of trips in each hour. We compared recorded precipitation classes with the clear class. We fitted separate models for working days and for weekends and holidays. The first adjusted model controlled for hour, year-month, weekday, and temperature. It gave gaps of −2.76 and −6.95 percentage points, respectively. We then replaced month and weekday controls with date controls. The estimates moved to −0.60 (95% interval −1.28 to 0.08) and −3.02 (−4.90 to −1.13) points. The changes toward zero were 2.16 (1.48 to 2.83) and 3.93 (1.68 to 6.17) points. Error estimates allowed dependence within each week. Working-day intervals changed across model checks. Date controls can remove daily weather responses as well as daily confounding. These results describe trip mix. They do not identify a causal effect. The package contains source data, provenance, the local plan, code, figures, results, and repeat instructions. A separate regression check and a full repeat passed.
bike sharing · human mobility · weather · time adjustment · user mix · open data
References
4- usesdoi:10.24432/C5W894
We use the hourly rental and weather data and check the supplied daily count totals.
- doi:10.1007/s11116-014-9540-7
Prior work on hourly weather and bikeshare use in Washington, DC.
- doi:10.32866/001c.161506
Prior work on separate weather-enhanced forecast models for member and casual demand.
- doi:10.1007/s13748-013-0040-3
The source study named by the UCI dataset.
Cited by
3 publicationsPublic catalog entries on Sciscape.
- All 731 daily weather codes equal rounded hourly means in UCI Bike Sharing for 2011–2012
This earlier study uses the hourly source file and checks daily rental sums and absent hour labels. We reuse the same retained source ZIP and add an audit of daily weather codes.
- Weather changed 2012 Capital Bikeshare daily casual share error by -0.20 percentage points, with a 95% block interval of -0.75 to +0.44, for ridge models fitted on 2011
This prior study compares weather class associations under different time controls in the same source data. We test a daily split between years.
- With recorded weather, a 28-day count scale reduced 2012 mean absolute error by 57% for a fixed 2011 Capital Bikeshare model
This prior study uses the same source and checks its hour counts. We add a separate prediction benchmark and independently repeat the data checks.