Experimental sampling was part of archaeology, but probably not part of other social sciences. The reason for this, as stated above, is that archaeology imported, with healthy skepticism, sampling theory from sociology and cultural geography. The sampling paradox and the archaeological research paradox both came in with the sociological tide, both paradoxes remaining under-appreciated to this day in archaeology. More research concerning the ‘archaeological research as cluster sampling’ paradox and its possible solution by jackknife and bootstrap subsampling is needed. Before adopting the sociolo-gists-geographers sampling as their own, archaeologists had to convince themselves with empirical proof that the theory worked on real archaeological data, as opposed to sociological or geographical data. Therefore, we stayed indoors, as armchair archaeologists, and compared fictive, hypothetical samples to total 100% archaeological populations based on our own previous, semi-good, old fashioned field work. Did experimental sampling work? Did it increase the use of probabilistic sampling in archaeology? The answer is probably yes. Was it necessary? Probably not, because the theory of statistics is so strong that our case studies are pale in comparison. Has experimental sampling contributed anything to archaeological knowledge or to sampling theory in general? Probably not.
Within the last 30 years since the late 1970s, probabilistic sampling has loosened its collar, so to speak, with the advent of model-based designs, such as reflective, bootstrap and jackknife sub-sampling, and of adaptive sampling. What used to be perceived as vast differences between probabilistic and nonprobabilistic sampling is evaporating into postmodern space. Rather than a traditional view of probability as an onerous straitjacket, Mueller and Orton, among others, have realized that sampling and statistics can be used creatively to enhance a project. One such use is the intellectual freedom to operationalize an anthropological concept in terms of the empirical reality of the archaeological record. One archaeologist may operationalize a concept one way, while another archaeologist operationalizes a concept another way. This is the accepted subjectivity of statistical inference as a logical tool; sampling and statistics also have their objective sides, as is commonly known. A second creative use is finding the proper match between the statistical test, the method of measurement, and the sampling scheme to fit the project’s research objectives. Thirdly, one can express one’s doubts (or conversely, one’s confidence) about a tested idea/hypothesis both quantitatively and graphically. Quantitatively, one can say that there is a 5% chance, for example, that the observed values between two samples will be greater than what we observed. Graphically, the critical value (or interval) can be plotted on the x-axis of a distribution curve and its relation to the mean and to the population’s dispersion can be expressed visually. And finally, the most powerful creative use is the ability to generalize from a small sample in one project’s data collection to a large target population and measure the probability that the generalization is correct.
See also: Statistics in Archaeology; Computer Simulation Modeling; Seasonality of Site Occupation.
World History
