Jeff Aa wrote at Thu, 19 Sep 2002 19:08:44 +0200: >> I liked your algorithm as it is quick and easy. >> But on the other hand the results aren't as good as possible. > > Actually, I do believe that the results are as good as possible, for any > distribution of numbers. 8-) consider that the first pass creates the > best possible approximation available by pairing the min and max. The > second pass merges the remaining values into the approximations, by > processing both the approximations and remaining numbers in sorted order > - ergo, it will produce the best possible result, irrespective of the > distribution of numbers.
Let's take an easy example: Numbers 1 8 27 64 125 216 343 512 729 Your algorithm will first group the mins/max => ( 1 729) ( 8 512) (27 343) and then assign the remaining numbers, so that the final result is ( 1 64 729) (avg 264,67) ( 8 125 512) (avg 215) (27 216 343) (avg 195,33) while the script I posted prints 1 8 729 avg 246 27 125 512 avg 221.333333333333 64 216 343 avg 207.666666666667 what seems to be really a bit better Greetings, Janek -- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]