(1) Develop model for data collection process (2) Adapt algorithm for map matching (3) Develop model for routes
- riders' differing rating criteria
- ride length
- time of day (for traffic and maybe other things)
- weather:
- Daily: avg. temp, mean wind speed, max gust speed
- Hourly: rainfall during ride, rainfall past 4 hours
- weather + ride length with linear predictors
- riders with random intercepts
- time of day effect with cyclic splines (by weekday/weekend)
- $y_i$ are ratings
- $X$ are ride predictors
- $j[i]$ is the rider of the $i$th ride
- $t_i$ is the start time of ride $i$ in hours
- $w_i$ is an indicator for weekend
- $s_{w_i}(t)$ are two cyclic cubic splines, for weekdays and weekends, w/ knots at every 3 hours.
- Model fit with gamm4 package in R
Photo by Richard Drdul. (https://www.flickr.com/photos/drdul/177247505)
The random intercepts were very informative for the model:
$\alpha_{j[i]}$ | $\beta X_i$ | $s_{w_i} (t_i)$ | AIC |
---|---|---|---|
X | X | X | 9096 |
X | X | 10755 | |
X | X | 9127 | |
X | X | 9153 |
- Maybe: Effect of cyclists' typical route absorbed by intercept
- This is a start towards a goal of building a better "Stress Map""
Future models should:
- Make use of the rider intercepts, time of day splines, and weather data
- Use more granular weather data
- Be aware of the correlation between rider intercepts, time of day, and route
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.