killbot2000
New member
- Joined
- Feb 5, 2016
- Messages
- 5
I'm working with a method called Kendall's Coefficient of Concordance. It allows you to get a measure of agreement between rankers for some items, e.g. movie reviewers for movies. You have a matrix with the reviewers along the rows and the movies in the columns where the elements represent the rankings. The result is a value between 0 and 1.
There's a paper discussing what to do in the case of missing data e.g. if a reviewer hasn't seen a movie. In this case you'll have zeroes in your matrix. https://www.researchgate.net/publication/4738093_The_coefficient_of_concordance_for_vague_data
However, one of the requirements is to have nondegenerated rankings. I'm not clear what this means. For instance, if I apply the method to my data, the measure goes above 1, which is nonsense. Can someone clarify what is meant by nondegenerated rankings? Thanks.
There's a paper discussing what to do in the case of missing data e.g. if a reviewer hasn't seen a movie. In this case you'll have zeroes in your matrix. https://www.researchgate.net/publication/4738093_The_coefficient_of_concordance_for_vague_data
However, one of the requirements is to have nondegenerated rankings. I'm not clear what this means. For instance, if I apply the method to my data, the measure goes above 1, which is nonsense. Can someone clarify what is meant by nondegenerated rankings? Thanks.
| 1 | 0 | 6 | 0 | 0 | 2 | 5 | 0 | 4 | 3 |
| 4 | 0 | 2 | 1 | 5 | 0 | 0 | 3 | 6 | 0 |
| 4 | 1 | 3 | 6 | 2 | 0 | 0 | 5 | 0 | 0 |
| 1 | 0 | 3 | 0 | 0 | 0 | 0 | 2 | 4 | 5 |
| 1 | 0 | 2 | 0 | 0 | 0 | 0 | 4 | 0 | 3 |
Last edited: