Uberdata: Rain's Effect on Uber Ridership

Man you folks are more harsh than peer-reviewers! :)

And here I was thinking to myself when I wrote the post, "people don't want to be drowning in data; they want something interesting based on the data". Dang. I was wrong!

On its own, that chart would be shady, you're right. For the purposes of the data I was trying to show (that weekend and weekday rides are quite similar), that was the best way to plot the data.

Of course if I take a linear function, and plot its integral, it will be exponential. By definition. I recognize that.

But that's why I also plotted the weekly averages farther down. You can see the non-linear growth very clearly in that chart. In fact, if I log-transform the average rides per week, the plot is quite linear (as would be expected for exponential growth). r(ln(rides)) = 0.83

Your comment about the slope since last November being "pretty linear" is misleading, too, though!

If you take an exponential function, and arbitrarily select any sub-segment of it, that sub-segment will always have a better linear fit than the linear fit for the whole function.

So by definition, mathematically, you're right. Our growth since November has been more linear than the overall history of our growth. Just as the first three months were also more linear. Or during any other arbitrary time period.

So I happily submit my revised manuscript for your reconsideration. I hope I have adequately and thoroughly addressed all of the reviewers concerns. Thank you.

Touché. I was definitely being technically unclear. My usage there was meant to indicate that the effect of the holidays on ridership was wildly higher than we had expected. Now that we've had a few holidays, we can begin to build that crazy surge into our models.

So yes, certainly the holidays themselves are predictable, but their effect on rides was not.

But now is. :)