Misleading averages and data visualization
In the course «Multivariate Analysis applied to Marketing» some students told me that the following diagram showed that sales are greater on Monday and Wednesday. How is such interpretation possible, after several months insisting on the importance of confidence intervals? How many bad decisions are made in businesses due to the mistake of taking into consideration average values instead of confidence intervals?
After a discussion about the issue with my LinkedIn contacts, we arrived to the conclusion that some best practices in data visualization could help to avoid such mistakes:
1) Vertical axis should start at cero, so visual perception improves even though viewers do not read labels.
2) Eliminate circles that represent averages. Just keep the segments that represent the confidence intervals.