Hi David, this sounds like a rather complicated problem to me. But perhaps I do not understand your real problem at hand. Could you try to be a little more specific, i.e. give a little background as where this problem comes from?
As I read it, you have data, that are corrupted by noise that is dependent on the magnitude of the data itself. I think it would be possible to come out with a model in principle. Do you know anything about how the noise depends on the data? Do you know anything about the noise? I personally would not try simply throwing away, some of the data completely (also this would destroy the original distribution), but use them (in a weighted way) because they still contain valuable information. regards, Roland . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
