Ranking Candidates

#1
If I have contestants complete 3 timed events each, what is the best way to rank each person based on their combined times (assuming each event has equal weight)?

Is it better to:

Add the times, compute an average and rank?
Just total all of the times and rank by total?
Figure a percentile rank for each time and then average the percentiles and rank?
Some other fancy technique?

Obviously, I'm no stats wiz....!

Thanks for anyones help!

John
 

Mean Joe

TS Contributor
#2
It depends on your definition of best.

Here's an example: Kris and Pat run 3 races (100m, 100m hurdle, 1500m).
Kris' times are: 10, 12, 500 (average = 174)
Pat's times are: 20, 30, 450 (average = 166.7)

Who's better? Kris won 2 of 3 races in half the time that Pat ran them, but in the longer run Kris was many more seconds slower (although not taking twice as long). In many cases (but not all) the average will tell you who is best all-around. But in this example it could be argued that Kris dominated two of the races.

There's a technique called Principal Components Analysis you may be interested in.
 
#3
Doesn't PCA just tell you which events are predictors of others? Either I don't udnrstand PCA (most likely) or I'm not sure of the application in this case.
 

vinux

Dark Knight
#4
One good measure is
Rank the each event and Sum or Average the Ranks by contestant ( since you mentioned equal weight assumption)
Then we can compare the contestants