OUTLIER TREATMENT IN DATA MERGING

Experience with a variety of diffraction data-reduction problems has l ed to several strategies for dealing with mismeasured outliers in mult iply measured data sets. Key features of the schemes employed currentl y include outlier identification based on the values y(median) = media n(\F-i\(2)), sigma(median) = median[sigma(\F-i\(2))], and \Delta\media n = median(\Delta(i)\) = median[\\F-i\(2)-median (\F-i\(2))\] in sampl es with i=1,2,...,n and n greater than or equal to 2 measurements; and robust/resistant averaging weights based on values of \z(i)\=\Delta(i )\/ max{sigma(median), \Delta\(median)[n/(n-1)](1/2). For outlier disc rimination or down-weighting, sample median values have the advantage of being much less outlier-based than sample mean values would be.