This week’s theme is sampling. Until the recent Stage 3 Meaningful Use rule goes into effect, making electronic health records (EHRs) at least capable of collecting sexual orientation and gender identity (SOGI) data, and people start actually collecting said data, nice, big probability samples of gender and sexual minority people are going to remain few and far between. So, we end up using non-probability sampling methods and try to take as much bias out of them as we can. Here’s me babbling about the various things we read on that topic for my class this week.
I did change my angle so that there is no exploding TARDIS in the background this time. Instead, you get to see my bulletin board with a map of Hartford and various shelters that are thumbtacked for my undergrad service learning students. Sound actually got worse, though. Trial and error continue!
Henry, G.T. (1990). Introduction and Sample Selection Approaches. In Henry, G.T. (1990). Practical Sampling. London: Sage.
Not a whole lot in the way of new info if you’re already familiar with basic sampling strategies, but a good intro if you’re not.
Sell, R.L., Silenzio, V.M.B. (2006). Lesbian, Gay, Bisexual, and Transgender Public Health Research. In Shankle, M.D. (Ed.). The Handbook of Lesbian, Gay, Bisexual, and Transgender Public Health: A Practitioner’s Guide to Service. Binghamton, NY: The Haworth Press, Inc.
Overview of sampling strategies common for this population (network, outcropping, advertising) touches on why they may be useful or the only available approaches. Also a helpful discussion on strategies for shaping questions to make members of stigmatized populations more comfortable with what is being asked.
Sell, R.L., Kates, J., Brodie, M. (2007). Use of a telephone screener to identify a probability sample of gays, lesbians and bisexuals. Journal of Homosexuality 53(4), 163-171.
Demonstrates one way to get a fairly large probability sample, using random digit dialing and oversampling metropolitan areas, then correcting for the oversampling in the statistical analysis. A major limitation, noted in the article, is that people may give negative responses out of fear of the consequences of being outed, though it was also noted that the resulting percentages were consistent with other studies on prevalence of gender minority persons. That may, of course, just indicate that the number of people willing to identify themselves had remained largely stable over time.
Heckathorn, D.D., Jeffri, J. (2003). Social Networks of Jazz Musicians. In Changing the Beat: A Study of the Worklife of Jazz Musicians, Volume III: Respondent-Driven Sampling: Survey Results by the Research Center for Arts and Culture, National Endowment for the Arts Research Division Report #43, pp. 48-61, Washington DC.
Demonstrates Respondent-Driven Sampling (RDS) with a population of artists. Mentions that for RDS to work, there has to be a communication pattern that connects the population. With gender and sexual minorities, the risk is that you miss severely closeted people who do not participate in any such communication pattern, and these are people who may have different health issues and outcomes than those who do participate. So it doesn’t quite have the generalizability of probability sampling but does result in a more representative sample than other non-probability methods. It’s worth noting that while jazz musicians are not a notably stigmatized group, the article mentions that this method was originally developed to improve HIV/AIDS research.