Particularly, the recent Monte Carlo simulations supporting these claims [27-33] are based on a misunderstood role of the

mean of the sample parameters.

If the

mean of the sample constructs after transferring the linguistic value to representative value, the calculation would be easier.

Sites 1,4&6 Sites 2,3&5 Sites 1,4&6 Variable Inside Outside Total Chlorides Number of paired samples 77

Mean of the sample logs 4.04 3.96(*) t=4.05 Antilogarithm of the mean of sample logs 10965 9120 * p|is less than~.05

(N) 50 50 50 50 50 50 A-60% 46% 40% 28% B-35% 46% 70% 48% 44% 68% C-0% 7% 30% 6% 56% 4% * This is the

mean of the sample as a whole.

The third, referred to as the "Biased Mean," was the

mean of the sample of patient ratings that resulted from differentially sampling patients as described above, where the most satisfied patients had a greater probability of being included.