Achieving a Critical Mass While Avoiding an Explosion: Letter-Cluster Sampling and Nominal Record Linkage

Abstract

Achieving a Critical Mass While Avoiding an Explosion: Letter-Cluster Sampling and Nominal Record Linkage It is now axiomatic that random sampling is advisable when entire populations, or even large groups, are to be analyzed; random sampling can reduce the largest data sets to more manageable proportions with only marginal and measurable losses in accuracy and reliability. The value of longitudinal analysis is also widely recognized. An analysis extended over time can reveal much that remains impervious to static assessments, particularly when attributes and behavioral patterns of groups of specific individuals can be compared at different points in time, thereby forming a panel. However, the formation of such panels from historical data involves nominal record linkage, and, unfortunately, nominal record linkage and random sampling are basically incompatible.1 The recent spate of discussions of nominal record linkage systems is indicative of the general utility of linkage for historical research, but the practical value of nominal record linkage is severely diminished by the inapplicability of random sampling. The problem raised by the incompatibility of these two techniques has not escaped the attention of those engaged in prosopographical studies involving record linkage.2 The problem is only too