From: "William H. Press" Date: 2005 April 29, 8:59:46 AM MDT To: kankel@mithra.physics.montana.edu Cc: Saul Teukolsky Subject: ks2d1s bug Charles, Sorry that we let your bug report fall into a crack two years ago! I've now posted it on the Forum at www.nr.com. Could you take a look and see if I've got it right? I did make a change to your proposed fix for quadct. Seems to me that your version would not count *at all* points that lie on the coordinate axes, and that it is better to count them, even if arbitrarily, so that the normalizations remain predictable. I agree that this is a case of measure zero for "general" data -- but people often have data that is discretized fairly coarsely. What do you think? I suppose randomizing them would be even better, but that seems beyond the scope of a simple recipe. Cheers, Bill Press