Now that I think about, I guess I kinda gave a bad example. For the values that might be duplicated, they would be integers which "should" be added together and then stored into the hash. So I guess there wouldn't be a value that has more than one value for particular record.
With this new info do you still think it's necessary to do a hash of hashes still?
Re^2: Assign 2D-Arrays to a hash using a key
It may not be necessary, if your data is reliable enough that you know there will always be the same number of values in the same order for each key. But I still like the hash-of-hashes because it tends to be easier to remember key names than index numbers. For instance, which of these makes it easier to recognize what it contains?