quote:
Originally posted by redstang281:
... the thing to do is get a sequence of dates and throw out those that are vastly anomalous. (Curtis et al)
Seems reasonable enough you gather a statistical population, find mean and sigma and apply an iterative method to eliminate statistical fliers.....
if you were to use 3 sigma limits you would be 97.5 percent sure that any point eliminated was not part of a gaussian distribution around the mean but a mis-measurement or a contaminated sample...
At least thats how I`d do it and they are smarter than me so I doubt they just pick results on a whim.....