Jaybear wrote:
The amicus brief was filed jointly filed by 1) the American Psychological Association; 2) the California Psychological Association; 3) the American Psychiatric Association; 4) the National Association of Social Workers; and 5) its California Chapter; 6) the American Medical Association; 7) the American Academy of Pediatrics; and 8) the American Psychoanalytic Association.
There is no rational, good faith debate over the question. Of course, that won't stop people like Hobo1512 from expressing a cynical belief that these organizations and leading experts in the field have conspired to set aside the best interest of children in order to advance the cause of gay rights.
The confident pronouncements from Jaybear and Chap, neither of whom has demonstrated the slightest intellectual competency in statistics, are precious. Professor Regnerus is absolutely correct when he writes:
Concern has arisen, however, about the methodological quality of many studies focusing on same-sex parents. In particular, most are based on non-random, non-representative data often employing small samples that do not allow for generalization to the larger population of gay and lesbian families (Nock, 2001; Perrin and Committee on Psychosocial Aspects of Child and Family Health, 2002; Redding, 2008). For instance, many published studies on the children of same-sex parents collect data from ‘‘snowball’’ or convenience samples (e.g., Bos et al., 2007; Brewaeys et al., 1997; Fulcher et al., 2008; Sirota,
2009; Vanfraussen et al., 2003). One notable example of this is the National Longitudinal Lesbian Family Study, analyses of which were prominently featured in the media in 2011 (e.g., Huffington Post, 2011). The NLLFS employs a convenience sample, recruited entirely by self-selection from announcements posted ‘‘at lesbian events, in women’s bookstores, and in lesbian newspapers’’ in Boston, Washington, and San Francisco. While I do not wish to downplay the significance of such a
longitudinal study—it is itself quite a feat—this sampling approach is a problem when the goal (or in this case, the practical result and conventional use of its findings) is to generalize to a population. All such samples are biased, often in unknown ways. As a formal sampling method, ‘‘snowball sampling is known to have some serious problems,’’ one expert asserts (Snijders, 1992, p. 59). Indeed, such samples are likely biased toward ‘‘inclusion of those who have many interrelationships with,
or are coupled to, a large number of other individuals’’ (Berg, 1988, p. 531). But apart from the knowledge of individuals’ inclusion probability, unbiased estimation is not possible.
Further, as Nock (2001) entreated, consider the convenience sample recruited from within organizations devoted to seeking rights for gays and lesbians, like the NLLFS sampling strategy. Suppose, for example, that the respondents have higher levels of education than comparable lesbians who do not frequent such events or bookstores, or who live elsewhere.
If such a sample is used for research purposes, then anything that is correlated with educational attainment—like better health, more deliberative parenting, and greater access to social capital and educational opportunities for children—will be biased. Any claims about a population based on a group that does not represent it will be distorted, since its sample
of lesbian parents is less diverse (given what is known about it) than a representative sample would reveal (Baumle
et al., 2009).
To compound the problem, results from nonprobability samples—from which meaningful statistics cannot be generated—are regularly compared with population-level samples of heterosexual parents, which no doubt are comprised of a blend of higher and lower quality parents. For example, Gartrell et al. (2011a,b) inquired about the sexual orientation and behavior of adolescents by comparing data from the National Survey of Family Growth (NSFG) with those in the snowball sample of
youth in the NLLFS. Comparing a population-based sample (the NSFG) to a select sample of youth from same-sex parents does not provide the statistical confidence demanded of good social science. Until now, this has been a primary way in which scholars have collected and evaluated data on same-sex parents. This is not to suggest that snowball samples are inherently
problematic as data-collection techniques, only that they are not adequate for making useful comparisons with samples that are entirely different with regard to selection characteristics. Snowball and various other types of convenience sampling are simply not widely generalizable or comparable to the population of interest as a whole. While researchers themselves commonly
commonly note this important limitation, it is often entirely lost in the translation and transmission of findings by the media to the public.
...
However, small sample sizes can contribute to ‘‘no differences’’ conclusions. It is not surprising that statistically-significant differences would not emerge in studies employing as few as 18 or 33 or 44 cases of respondents with same-sex parents, respectively (Fulcher et al., 2008; Golombok et al., 2003; Wainright and Patterson, 2006). Even analyzing matched samples, as a variety of studies have done, fails to mitigate the challenge of locating statistically-significant differences when the sample size is small. This is a concern in all of social science, but one that is doubly important when there may be motivation to confirm the null hypothesis (that is, that there are in fact no statistically-significant differences between groups).
...
However, the meta-analysis reinforces the profound importance of who is doing the reporting—nearly always volunteers for small studies on a group whose claims about documentable parenting successes are very relevant in recent legislative and judicial debates over rights and legal statuses. Tasker (2010, p. 36) suggests caution:
'Parental self-report, of course, may be biased. It is plausible to argue that, in a prejudiced social climate, lesbian and gay parents may have more at stake in presenting a positive picture. . ..Future studies need to consider using additional sophisticated measures to rule out potential biases. . .'
Professor Regnerus is absolutely correct in these criticisms. (I similarly criticized Gartrell's study before the Regnerus paper was published.) Jaybear, Chap, and others uncritical acceptance of the studies in cited in the APA brief, which are junk, speaks volumes.