User Control versus Randomisation in Geographical Probability Sampling: A Compromise Solution Using Controlled Sampling

Probability samples based on some variant of random selection are often used in geography to obtain data suitable for analysis by inductive statistics. However, because geographical populations are frequently skew or irregularly distributed such samples may not always possess desirable spatial properties. Ideally, the researcher would wish to exert control over the composition of the sample while continuing to use random selection. The purpose of this paper is to present a method which can facilitate this—controlled sampling—and illustrate it using retailing data collected for the Cardiff Consumer Panel Survey of 1982.