Possible Duplicate:
Averaging correlation values
I have a number of users, and for each user I have 2 ordered lists of items (one is from observation, one is generated by an algorithm). For a given user, I can use the Spearman correlation to gauge how well my algorithm is performing. However, I'd like to evaluate the algorithm on the entire population. Is it valid to average the Spearman correlations? If not, how can I generate a single value that aggregates a large number of the correlations?
Also, I'm a beginner in this field -- please let me know if I'm going down the wrong path totally here. Any links to articles/papers about this sort of thing are much appreciated.
