An Empirical Investigation of Sampling Strategies in Marital Research

Abstract

Critical analysis of the methods employed in marital research has often led progress in understanding marriage. For example, skepticism over the validity of self-report methods (e.g., Raush, Barry, Hertel, & Swain, 1974) led observational coding of marital interactions and greater understanding of marital communication (see Baucom & Adams, 1987). Similarly, concerns about the limitations of cross-sectional designs (e.g., Huston & Levinger, 1978) led longitudinal studies and clarification of how marriages change over time (see Bradbury & Karney, 1993). As a consequence of these and other developments, marital researchers are in an unprecedented position investigate diverse and complex issues regarding how marriages function. In a field marked by such methodological scrutiny, it is surprising that little attention has been devoted understanding and refining the procedures used recruit research participants. For example, there appears be no consensus in marital research on the most appropriate means of sampling married couples. A recent review of longitudinal studies of marriage found that 27% of this research examined samples of convenience (e.g., students, patients), 23% analyzed samples obtained for purposes other than studying marriage, 22% solicited volunteers through newspapers and other media, 18% used random sampling, and 12% recruited through publicly available marriage license records (Karney & Bradbury, 1995). Despite this variety, and despite warnings by Kitson et al. (1982) over a decade ago that, to make further improvements in the quality of research on marriage and the family, an adequate body of knowledge about sampling methods used in earlier research is needed (p. 966), we know of no research that has examined the implications of these different sampling techniques for the samples obtained in marital research. In the continued absence of such information, attempts by researchers draw from and expand upon previous research on marriage are likely be hindered in three ways. First, lack of data on the comparability of different sampling techniques renders difficult any synthesis of studies employing different techniques. Second, researchers lack empirical guidelines for choosing sampling techniques that may be most appropriate for specific questions. Finally, in view of the fact that most longitudinal studies of marriage are conducted on middle-class Caucasians (Karney & Bradbury, 1995), and the fact that overlooked segments of the population (e.g., couples who are poorer, with less education, and from ethnic minority backgrounds) appear face unique challenges in marriage (e.g., Mott & Moore, 1979), neglect of sampling issues may be resulting in the exclusion of a broad range of couples from marital research. As a result, even methodologically ambitious studies and the theories which they give rise may provide only a limited understanding of marital phenomena. The present article describes two studies designed address these shortcomings in the literature by examining the effects of different sampling procedures on the samples obtained in marital research. In both studies, we focus on sampling procedures used recruit couples for longitudinal studies because of the growing emphasis placed on such designs and the great cost involved in collecting longitudinal data. Special attention is paid the issues of sampling frame and sample coverage. SAMPLING FRAME Ideally, sampling procedures should be chosen draw participants from the population most appropriate testing a given hypothesis. With several common sampling techniques, however, the population from which the sample is drawn, or the sampling frame, is difficult determine. For example, studies soliciting couples through newspaper advertisements or snowball sampling (i.e., asking couples who participate recruit other couples) cannot determine the characteristics of the population exposed the solicitation. …