APPENDIX 2 DUPLICATES EPA (1977c) describes a technique for estimating the sampling error for individual values based on the results from duplicates. The subject technique is based on assuming the data can be fit by a lognormal distribution. Since the data treatments in this paper have been based on the assumptions of a normal distribution, here. this assumption will be applied The variance for the individual duplicates S? is (0.886R)2. Where R is the absolute difference between a pair of duplicates and 0.886 is a parameter for estimating the standard deviation based on the range of two values. Snedecor and Cochran (1967) note this is a statistically efficient technique for small sample sizes. The population variance for a group of variances based on duplicates is equal to the average variance. If the individual variances are not all for duplicates (e.g., triplicates and duplicates, etc.), the values have to be weighted by their respective n's. The estimates of the respective variances are given in the columm on the right in Table 2-1. The estimate of the population (average) variance is 4491 and the estimated standard deviation is 67 fCi/g. The estimated coefficient of variation is 29 percent (based on the average of the duplicates--230 fCi/g). If the lognormal distribution had been assumed, the estimated geometric standard deviation would be 1.53. 674