- #1

- 1

- 0

I tried the follow procedure:

-Sort the Z1/Z2

-Rank them and store the rank on a new column, i

-perform median rank (herd-Johnson) i/n+1 where n is the sample size

-perform inverse cumulative probability on the median rank column

-plot the z1/z2 vs inverse cumulative probability

What i get is near the location, the data are linear while the deviation is serious at either extreme ends. any suggestion? or references?