Interesting. The linked article seems to suggest that part of the differences are due to different companies having the same DNA data about the twins but different reference data sets (for what is "Balkan DNA", for example).
A good reminder that in many statistical analyses, there is not only uncertainness on the measure but also on the tool or on the reference used to evaluate the measure:
For example, when estimating how many goals will Cristiano Ronaldo score next year, there are 2 sources of uncertainty:
- His presumed skill might not match how many goals he'll score in the future
- How many goals he scored in the past might not have matched his skill.
- His presumed skill might not match how many goals he'll score in the future
- How many goals he scored in the past might not have matched his skill.
