any formal cost /benefit analyses made regarding these objectives. Most 

 inspections were done if there was suspicion of damage or on a periodic 

 schedule that was "based upon past experience and the availability of 

 funds." There were no inspection programs that used any statistical 

 sampling as a projection of facility condition. 



The lack of criteria and procedures from commercial sources and 

 concerns about the adequacy of current Navy inspection procedures 

 prompted the establishment of the project at NCEL to develop criteria 

 for preliminary inspection of waterfront facilities. The main emphases 

 of this project were threefold: 



(1) To develop procedures based on statistical sampling theory for 

 use in planning and assessing the results of underwater 

 inspections 



(2) To validate the theory using data from previous inspections 

 and conducting special inspections to test the procedures for 

 selection of data samples 



(3) To develop the criteria and procedures in a format easily 

 implemented by field personnel 



This report addresses these three areas. 



STATISTICAL SAMPLING THEORY 



There are two primary methods for establishing a sampling criterion: 

 (1) judgment sampling and (2) statistical sampling. In judgment sampling, 

 a subjective decision is made regarding the criteria for selecting samples 

 from a population. This is commonly used whenever there is expert knowl- 

 edge about the population being sampled and the investigator feels 

 "comfortable" with the choice made based on past experience with investi- 

 gations of this population. This method of sampling offers no guidelines 

 for determining the sample size nor any recognized method for the pro- 

 jection of the results of the sample onto the entire population. 



In the second sampling technique, called statistical sampling, the 

 sample is selected by a specified method of random sampling. The advan- 

 tage of statistical sampling is that it permits the objective measurement 

 of the variability of the sample and prescribes scientific methods to 

 project the results to the target population. Using the statistical 

 sampling method provides a mathematical rationale for evaluating the 

 tradeoffs between the cost of collecting the data and the selection of a 

 confidence level and a specified accuracy. 



Two major types of sampling plans, sampling by variables and sampling 

 by attributes, are discussed in the remainder of this section. Sampling 

 by variables occurs when a quality or parameter of each element of the 

 population can be measured using either a continuous or discrete scale. 

 These data are then analyzed to determine an average or mean value of the 

 parameter, as well as other statistical indicators of the distribution of 

 the data. Examples of variable sampling include measurements of the 



