Reviews sampling methods used in surveys: simple random sampling, systematic sampling, stratification, cluster and multi-stage sampling, sampling with probability proportional to size, two-phase sampling, replicated sampling, panel designs, and non-probability sampling. Kalton discusses issues of practical implementation, including frame problems and non-response, and gives examples of sample designs for a national face-to-face interview survey and for a telephone survey. He also treats the use of weights in survey analysis, the computation of sampling errors with complex sampling designs, and the determination of sample size.
Probability sampling avoids selection bias by giving each element on the sampling frame a known and nonzero probability of selection. The methods for dealing with frame problems described in the last section were developed to help to eliminate, or at least reduce, biases resulting from frame deficiencies. Given a good frame, a probability sample of the population may be drawn, but there still remains the need to collect the survey data from the sampled elements. Failure to collect the survey data from some sampled elements, or nonresponse, is a major survey problem that seems to have grown in recent years as the public has become less willing to participate in surveys (see, for instance, Steeh, 1981).
The cause of concern about nonresponse is the risk that ...