Our python notebook can be found at this link.
Conclusions -
Using the two-sample t-test, we found evidence, with a very high degree of significance, that Subscriber trip duration is shorter than Customer trip duration.
One weakness is Citibike's payment structure, which may artificially skew the distributions of our datasets. Customers pay a flat fee for the day, but must pay $4 more for each additional fifteen minutes after a 30 minute ride. Subscribers pay a flat monthly fee, but must pay an additional $2.50 for each extra fifteen minutes after a 45 minute ride. These incongruent payment structures may cause the data to have discontinuities at the 30 minute and 45 minute marks for Customers and Subscribers respectively.
A strength of our analysis is a clean-cut and simple analysis between two independent samples. The result is highly statistically significant and follows a simple logic.