Let's say I have two ranking list of homogeneous items. The rank list is prepared based on two different parameters, to find out the best items. Now is there any methodology that is helpful to do this. I have heard about some methods like, Rank Correlation method, Wilkinson Signed Rank List, Paired T Test etc etc, but don't know their applicability with varied data types. Can somebody help in this regard.
Hi Kalita
To rank items based on your preference is a good way to find the best compromise. However, comparing two ranking is difficult to say which is the best if you don’t regard your preferences in that comparison.
If you want to rank items based in your preferences there are some methods you can use: ELECTRE III, PROMETHEE, AHP.
You can use any comparison method between rankings to find out how similar or how different they are. Example: kendall correlation is simple. A weighted kendall version penalizes stronger top differences between ranking (rank reversal) than bottom differences.
I share you the paper of weighted kendall version, you can find a illustrative example about how to compare two rankings.
Any comment, let me know.
Good lucky.
Thanks Pavel Anselmo Alvarez Carrillo , it seems from your description that Kendall correlation have that feature which I am looking for. I will look into your paper, and contact you for further clarification. Thank you for your effort.
And also Pavel Anselmo Alvarez Carrillo, can I check agreement level between the two rank list with Kendall Correlation?
Yes, you can. It is natural to compare both ranking with Kendall Correlation approached as agreement level. However, it takes the same, rank reversal (reversed pairs) occurring on the top or bottom of the rank. It means that a rank reversal in bottom is same penalized as rank reversal in the top on rank. If it is not a problem in your research, Kendall Correlation could be good method to use.
Feel free to contact me if you want to discuss anything.
- Badges
- Science topic