"Zone" wrote in message
ups.com...


The problem
is that the data is very "raw" -- about 12,000 datapoints ranging from
0.22 to over 100,000. You probably know that this is the kind of mess
one sometimes gets from clients who don't really understand the ratios
that are being asked for.

Hi Zone,
if you want to use the 2.5 StDev rule for removing outliers than your data
should be normally distributed.Ratios are known to be notoriously skewed.

1) Do a transform ; a log- or arcsin transformation could be usefull
2) Do a Kolmogorov-Smirnov test to see if you now have a normal distribution
3) Do the outlier test on the new data

Use statistics software if you can , SAS , Statistica , R , S
hth
Gys