These confidence limits (the probable error) are shown graphically in Figure 10 for values of ¢ of 0.1 to Lil wet 0.2 increments and in tabular form for the 0.7 value of c in Table 4. Inspection of Figure 10 will show the effect of increasing the number of samples or of assuming or having a different value of the coefficient of variation. For any particular value of the coefficient of variation, the effect of inecreactin= the sample numberc on the 7eft side cf the curve (less than 30 samples) has a much more positive effect on the precision than increasing the same number on the right side (greater than 30 samples). For example, using a coefficient of variation of 0.7, thirty samples produce an expected Probable Error of + 27h. An increase of precision to + 15% would require eighty samples, and to + 10%, one hundred eighty samples. The law of diminishing returns is obvious. Further inspection of the figure will illustrate the necessary compromises considered in establishing appropriate confidence limits. The actual confidence limits were established according to the information required. These limits then indicated the number of sample collection locations required. The actual numbers of sample locations by island and strata group, with the indicated assumed confidence limits and assumed mean are listed in Table 5. did not differ too greatly from the designed values, They