Expanding Statistical Frontiers in Sexual Science: Taxometric, Invariance, and Equivalence Testing

Abstract

Sexual scientists must choose from among myriad methodological and analytical approaches when investigating their research questions. How can scholars learn whether sexualities are discrete or continuous? How is sexuality constructed? And to what extent are sexuality-related groups similar to or different from one another? Though commonplace, quantitative attempts at addressing these research questions require users to possess an increasingly deep repertoire of statistical knowledge and programming skills. Recently developed open-source software offers powerful yet accessible capacity to researchers wishing to perform strong quantitative tests. Taking advantage of these new statistical opportunities will require sexual scientists to become familiar with new analyses, including taxometric analysis, tests of measurement variability and differential item functioning, and equivalence testing. In the current article, I discuss each of these analyses, providing conceptual and historical overviews. I also address common misunderstandings for each analysis that may discourage researchers from implementing them. Finally, I describe current best practices when using each analysis, providing reproducible coding examples and interpretations along the way, in an attempt to reduce barriers to the uptake of these analyses. By aspiring to explore these new statistical frontiers in sexual science, sexuality researchers will be better positioned to test their substantive theories of interest.

 Funding and Acknowledgment

I wish to thank Drs. Robin Milhausen, Monica Biernat, and Emily Impett for consistently empowering me to pursue and capitalize on my data-analysis-related training; Shari Blumenstock, Robyn Kilshaw, Chris Quinn-Nilas, and Dr. Shayna Skakoon-Sparling for their valuable feedback on a previous draft of this manuscript; Sally (Yue) Lin for introducing me to the concept of knowledge brokering; Dr. J. Ruscio and two anonymous reviewers for their helpful feedback; and Dr. Sari van Anders for her guidance, trust, and encouragement throughout the entire process of writing this review. This research was supported by an SSHRC Insight Development Grant.

Notes

¹ There is some flexibility on this point, as current methods enable researchers to use categorical indicators for estimating common factor models (e.g., Rhemtulla, Brosseau-Liard, & Savalei, 2012) and continuous indicators for IRT models (e.g., Masters, 1982) if they wish.