Exposing Biased Samples: Derivation of the Sample-subdivision Method

In most research programmes, the initial sample is randomly selected, whereas the final sample is self-selected and often biased. This paper derives a technique that determines whether or not the final sample is biased. Specifically, the initial sample is randomly subdivided into groups. For each group, the statistic of interest and the response rate is calculated. A relationship between the statistic of interest and the response rate reflects a biased sample. Conversely, the absence of any relationship reflects an unbiased sample. The advantages of this procedure vis-a-vis past practices are underscored.