Supplement - Page 2. a. 138 .01 2.35 05 10 100(1-a)% Upper Limit on True Mean 20 pCi/g 1.66 1.29 19 18 Since the decision to cleanup may be a function more of extreme values rather than average concentrations the next section considers upper limits on percentiles. II. Nonparametric Confidence Limits on Percentiles Using "Practical Nonparametric Statistics" by W. J. Conover, John Wiley, 1971, page 111, we compute upper one-sided confidence limits: The probability is l-a that p percent of the soil concentrations for the area from which samples were collected are less than or equal to X. Estimated values of X for various values of p and a for the data from Janet are: P a X(pCi/g) -50, .05 iB] -50 25 10 .90 -01 . 90 10 507 -50 .90 .01 13 -10 51 .05 46 -90 .25 4] -95 -01 120 95 .95 .10 25 57 52 .95 Interpretation: (median = 9.8 pCi/g) 1] (90°" 4] 05 67 . _ For p = .90 and a = .05 we state: (95° th th percentile = 37 pCi/g) percentile = 46 pCi/g) We are 95% sure that 90% of the soil concentrations on the island are < 46 pCi/gm. TThese values of X for a = .01, .05, and .10 when p = .50 are nonparametric equivalents of the 100(1-a)% upper limits on the median computed in Part I above. The upper limits (X) obtained here do not require any assumption about the distribution of the observations. Note that these limits are consequently somewhat higher than the corresponding limits in Part I.