Some Tips to Read the Results
If we don't know how to read the results, our analysis will be false. To prevent this, let's look at some tips.
Prefer Percentiles over Average
Average is used a lot because it's easy to understand and calculate.
But mean and median values do not correctly reflect the user experience:
- They tend to hide outliers.
- They distort reality.
- They hide the distribution of the data.
Different datasets can give you the same average as we can see in Anscombe's quartet (https://en.wikipedia.org/wiki/Anscombe's_quartet) or in this publication (https://www.autodeskresearch.com/publications/samestats).
In these two graphs, the averages are the same and equal to 5.578947368:
Figure 12.15: Average = 5.578947368
Figure 12.16: The same average as the previous graph (5.578947368)
One solution to solve these problems is to use percentiles (https://en.wikipedia.org/wiki/Percentile).
The XXth percentile (pXX) is the value...
