We started reading Active Statistics by Andrew Gelman. It has very nice pedagogy that suits online reading. We decided that every week we'd read the chapter prescribed and the real story Gelman worked upon. We would then discuss the 'drills' and other classroom discussion problems given over the call. There are also computer demonstrations and assignments that readers can practice.
This week's story was how Wikipedia messed up randomisation in a simple A/B experiment. Wikipedia once found that having a square donation box reduced the number of donations by 19% compared to a rounded donation box.
Obviously, it sounds weird. The difference is not even noticeable. How come this square box has such a negative impact? It was not a small experiment as well. The sample size was 2 Million!
After digging into the data, Gelmnam et al., found that Wikipedia messed up with randomisation. For some reason, the treatment group (square corners) received the donation box at night, when generally people don't donate. As a result of this messed up randomisation, they saw a huge negative effect on the square corners box. Kartik mentioned that having an eye on the "data generation process" will help us avoid such errors.
For the next week
As per the plan mentioned in Active Statistics,
3. (Optional) assignments, computer demonstrations from Chapter-3 (Class-2)
The call will be on October 18th, Saturday, at 12-1PM , unless changed. Please RSVP to the GMeet. I've sent the recurring invite
Please add anything I missed or misrepresented
Best,