2017-09-14

/Attendees

Minutes

Ran Cory's bar-chart versus pie-chart experiment (available at http://piechartssuck.com/) with 3 participants, and got:

• Bar-chart mean error: 2.133
• Pie-chart mean error: 1.733
• P-value: 0.50

Slightly surprising to swing the way of pie-chart estimating being more accurate, but the p-value indicates that the effect is small enough that we need more participants to reliably detect it.

Talked about whether the tick marks on bar-charts are biasing the experiment, and alternative ways to present the charts (something is needed to show the user where "100%" is for a bar-chart).

Read Cory's code. Discovered some tracking of subjects/observations may have been wrong, leading to a bad p-value. Cory's going to fix up the code and try this again later. However, the mean errors should be correct.

Next Time

• NEXT_MEETING_DATE

• Be ready to discuss Chapter XXX of AnIntroductionToStatisticalLearning