ranking multiple lists

zed

Legend
Im wanting an (possible) answer for the following question (of course theres not an exact answer but I'm OK with something reasonable)
http://stats.stackexchange.com/questions/89012/aggregate-ranking-from-multiple-ranked-lists
the social choice answer is not acceptable, btw.

The problem of course is that not everyone has tried every sort of fruit, if they had then the solution would be easy.
The more I think about it, the more I see you can't really come up with something satisfactory, and what will happen is the more commonly tried fruits eg Apples will come to the top because everyones tried them in place of the more exotic (in the west) eg Rambutan's or whatever.

OK 3 fruit choices isnt really enough, say increase it to 10, & have 10 ppl choosing their personal preferences.
10 ppl have banana's at between 8-10 on their list's
4 ppl have rambutan's between 1-3 on their lists (it doesnt appear on the other 6 lists, perhaps 2 tried it and hated it but the other 4 havent even tried it)

Looking at that I would put rambutans's higher, based on average list placement but I feel you've also gotta reward banana a bit extra for just being on more ppl's lists ( i.e. I assume everyone has tried more than 10 different fruits).
Does this make sense somewhat? Sorry if unclear
Any Ideas?, cheers zed
 
Id say you award a score eg 10 points for 1st, 9 for 2nd ect (assuming 10 fruits) and just add the scores
Its a pure data question no need to be thinking maybe only 4 people have tasted a kiwi fruit so I have to take that into account
In short your over thinking the question.
 
Id say you award a score eg 10 points for 1st, 9 for 2nd ect (assuming 10 fruits) and just add the scores
dude the major flaw in that is the most commonly eaten fruit rises to the top, this is like the IMDB film rankings, only well known films rise to the top (A),
Im wanting something closer to the ideal (i.e. if everyone tasted all fruits)(B)

I'm wanting a formula between A & B (but closer to B)
 
the most commonly eaten fruit rises to the top
Isnt that the point
If you want them rated by some other criteria you havnt explained what it is
and you cant take into account fruit that rates low because the questionee hasn't tasted it because you have no data on that.

you could massively oversample and only get stats from those questionaires that contain all the fruits you want to consider then for example take 100 of them at random and get your results from those.
If you want to consider all fruits there are thousands of them, you've got no chance.
 
Last edited:
Isnt that the point
no, rated by perceived quality is the answer (otherwise you get the answer the best resturant is macdonalds), which we all know is not the case (unless you talk money)
If you want them rated by some other criteria you havnt explained what it is
well that sounds like the 'social choice' answer

I can come up with an equation but Im thinking obviously others have done this before, in politics/sports/fruit whatever thus Im asking
hmm thinking a bit, perhaps take into consideration the chance X has eaten the fruit eg apples = 97%, rambutans = 20% (though with low percentage fruit 5% and less this will throw off the answer) but yeah thats a valid thought
 
One option is to use relative ranking. E.g., only store if the same person tried banana and kiwi ranked kiwi higher than banana. You can improve reliability later by setting a data threshold (needs more than x or x% of votes).
 
no, rated by perceived quality
How do you define the perceived quality of a fruit ?
Do you intent to have a survey where the question is "what fruit has the best quality" or "list the fruits you've eaten by quality"
would'nt people interpret that as "what fruit is your favourite" and dont people pick a favourite based on taste ? afaik no one picks their favourite fruit by convenience or any other criteria.
you need to tell us what question you are asking
because it looks to me like the problem is not how to order the data it's the method you use to get the data
 
Back
Top