1 74 TRANSURANIC ELEMENTS IN THE ENVIRONMENT 



of treatments to his experimental material. Studies, by random sampling, of uncontrolled 

 phenomena are not experiments in this sense. 



Perhaps it is advisable to digress here to make a more detailed dis'tinction between a 

 sampling design and an experimental design. Consider a study of the effect of, say, four 

 levels of concentration of some chemical substance on a specific kind of plant. A suitable 

 experimental design might use, say, 25 individually potted plants kept in a growth 

 chamber. Each of 5 randomly selected pots would be treated with the same 

 concentration, and one lot of 5 plants would be a control. Another randomization would 

 be used in assigning the 25 pots to places in the growth chamber. There would be thus 5 

 replicates of each type of treatment (concentration of chemical) and 5 control pots. 



In contrast, a sample survey design might be used to study the concentration of, say, 

 plutonium in a natural population of plants in the vicinity of a nuclear test site. Here we 

 can only observe, by sampling, the results of events over which we usually have no 

 control. There are no true replicates in the experimental sense. It is now fairly common 

 practice, however, to speak of "replicate samples," and we would only stress the need to 

 use the word "sample" to be precise. The investigator cannot control the physical 

 relationships involved; he has to take specimens from an existing population of plants in 

 the positions in which they occur. He can, of course, control the sampling process so that 

 he obtains several individuals of the same species growing rouglily the same distance from 

 ground zero and so on. 



Since science is commonly thought of as being practically synonymous with the 

 experimental method, many people prefer to regard observations taken on some 

 uncontrolled process as experiments. Such a view is largely immaterial and irrelevant 

 insofar as the mathematical and computational aspects of statistical methods are 

 concerned. Only when we begin to draw inferences from analyses of the data does the 

 real distinction between "experiment" and "observation" become apparent; i.e., in a true 

 experiment we can use rather homogeneous material and, by randomization, ensure that 

 any effects due to position in the growth chamber, genetic factors, etc., are reflected by 

 the error term in the statistical analysis. Without this element of deliberate control of the 

 experiment, there is no assurance that unknown extraneous factors will not also influence 

 the factors under study. Hence the aiialysis of data obtained by sampling is a rather more 

 hazardous affair. However, exactly the same statistical analyses and an identical 

 mathematical model can be used for either an experiment or a sample survey. 



For the present, many of the immediate needs for statistical guidance in transuranic 

 research programs can best be served from the sampling point of view. Many textbooks 

 and experienced statistical practitioners are available to aid in the design and analysis of 

 experiments. We are also interested in the design of experiments, but we have elected to 

 concentrate on sampling in this chapter. We will not try to be explicit as to the role of 

 source terms, but a long list of sources must, of course, be considered; e.g. ( 1) worldwide 

 fallout from nuclear events; (2) localized fallout from nuclear events; (3) localized 

 dispersion without a nuclear event, such as safety tests; (4) stack releases; (5) liquid 

 effluents; (6) accidents involving nuclear weapons; (7)burnup of SNAP devices; 

 (8) damage to other power sources; and (9) various kinds of eventualities concerning 

 stored wastes, transport, etc. 



We will not attempt to deal with studies of the biological effects of exposures to 

 various substances. We have discussed the statistical problems in assessing the effects of 

 low-level, chronic pollutants elsewhere (Eberhardt, 1975a) and can only note here that 

 they are even more perplexing than those associated with sampling alone. 



