8. Select a random sampling technique. Systematic sampling is 

 preferred over simple random sampling. For the type of facilities in 

 the underwater inspection program with the extent of redundancy inherent 

 in the pier structures, there will be a gain in the precision of the 

 samples. The gain in precision over simple random sampling results from 

 the natural stratification of pier structures, particularly the charac- 

 teristic bent and pile design. Systematic sampling ensures that a more 

 "even" distribution of pile samples is realized, thus enlarging the 

 sample coverage over the pier. 



It was found that the pier plan drawings can be used for systematic 

 pile selection with relative ease. For example, to select 30 samples 

 from a pier that has 2,000 piles, the total number of piles (excluding 

 the fender and perimeter piles, dolphins, etc.) is divided by the 

 required sample size. For this example, 2,000/30 results in a skip 

 interval of 66. Next, a random starting number in the first skip inter- 

 val (between 1 and 66) must be selected. A skip interval of 66 is used 

 to find the next pile to sample. In this way, only one random number 

 needs to be selected from a table of random numbers (Appendix B) or 

 using a random number generator available on many hand-held calculators. 

 The remaining samples are selected by systematically continuing along 

 the rows or bents using the skip interval to determine the spacing 

 between samples. This process is repeated until all the piles for this 

 sample are identified. 



Piles can be selected by proceeding lengthwise along the pier or 

 along each bent across the width of the pier. Any pattern is acceptable 

 as long as the skip interval is not a multiple of the number of piles 

 along the direction selected, in either length or width. 



During the inspection, if a pile to be sampled is not accessible or 

 the drawing is incorrect, choose another pile nearest the sample pile. 

 This will tend to add to the randomness of the sample and is quite 

 acceptable for analysis. Any changes from the original sampling plan 

 should be noted to assure that the pile plan used in subsequent inspec- 

 tions is correct. 



9. Collect samples. The data for each variable and attribute to 

 be measured in the structure are collected from the predefined sample 

 elements. The techniques and equipment used to collect the data must be 

 at least as accurate as the required precision of the statistical analysis, 

 It is essential that where variable data are required, the actual measure- 

 ment is reported rather than a subjective assessment such as "original 

 thickness" or "no damage." 



10. Analyze sample data. The initial analysis of the sample data 

 should be conducted on site to confirm that sufficient data have been 

 collected to meet the accuracy and confidence level requirements estab- 

 lished prior to the inspection. This analysis should be conducted in 

 two phases : 



(1) Confirm that the number of samples is adequate to estimate 

 the mean and standard deviation of the population. 



41 



