Material Accuracy (%) 



Timber 10 



Steel 4 



Mission Confidence (%) 



Nominal 90 



Moderate 95 



Critical 98 



6. Use the formula for estimating the mean to obtain the sample 

 size, if maintenance is based upon a critical measurement parameter, 

 such as effective diameter or cross-sectional area. The STATISTICAL 

 SAMPLING THEORY section of this report presents equations that can be 

 used for sample sizes based on either the expected standard deviation of 

 pile diameters or upon the expected coefficient of variation. Because 

 of the time required to take additional samples, it is recommended that 

 a minimum of 30 samples be taken during the initial phase of the inspec- 

 tion. 



The recommended minimum sample size of 30 elements proved sufficient 

 to obtain an estimate of the mean diameter and percentage of piles within 

 specific diameter intervals for a desired confidence level. Using this 

 sample size, an estimate of the percentage of piles that fall into the four 

 maintenance categories can be determined. However, if the lower limit of 

 the confidence interval for the percentage of piles in good condition or 

 in need of repair is less than 80 percent, then additional samples are 

 required . 



7. For attribute (proportional) sampling, base the initial number of 

 samples on the expected or historical proportion for the specific attri- 

 bute to be tabulated (see Sampling by Attributes section). At St. Helena, 

 the attribute to be determined was the percentage of pilings with borer 

 damage. Using a confidence level of 95 percent, an accuracy of 10 percent, 

 and an expected attribute proportion of 20 percent, the number of samples 

 needed is 81. 



The maximum sample size requirement for attribute estimation occurs 

 when the expected proportion is 50 percent of the population. Unlike 

 variable sampling, the number of samples needed for proportional sampling 

 for a given confidence level and accuracy can be determined by using the 

 most conservative expected proportion value, that of 50 percent. Assuming 

 a confidence level of 95 percent and an accuracy of 10 percent, the maximum 

 number of samples required is 97. 



Attribute sampling can usually be conducted during the Level I phase 

 of the inspection since measurement values are not required. Proportional 

 sampling requires only a binary value (usually a Yes or No) tabulation. If, 

 however, the extent of marine growth fouling prevents an accurate determina- 

 tion of the. attribute, then this type of sample will have to be obtained as 

 part of the Level II inspection. 



40 



