Ireland recovering from Theocracy.

[Sam]

It is entirely dishonest to say of opposing studies that "we discuss the handful of recent studies reporting that children fare worse on any measure of child well-being (Allen et al. 2013; Goldberg et al. 2011; Gartrell et al. 2011; Regnerus 2012a, b), and each has shortcomings making broad generalizations impossible" while not noting that the same applies to most of the studies cited explicitly or implicitly ("consensus", "40+ studies").

That depends entirely on the shortcomings. Vaguely recalling Regnerus, the shortcomings have nothing to do with sample-size problems like the NLLFS; the methodology was shot and the definition of a "same-sex parent" included anyone who admitted to having a homosexual encounter at some point ... such that Ted Haggard's kids would qualify as "children of a same-sex parent" in Regnerus' study. I'll have to look that one up again to remember the specifics but the point remains: we have to know what the shortcomings are before determining whether there's any hypocrisy.

So long as the NLLFS and gold-standard studies all point in the same direction, there's nothing unethical about saying that the broad consensus supports the hypothesis that children of same-sex partners have developmental parity with children of opposite sex partners. Even if the weaker studies have those shortcomings, they are consilient with nationally representative studies which do not and the statement is therefore supported.

[Sam]

Ah, okay, I misread your earlier statement. I would, however, like to know which of these are the case.

On the other hand we are, I hope, agreed that studies that use sampling from the NLLFS but make the invalid comparison to the national average have little to no evidentiary weight and that any consensus founded on these is bunkum?

Yes, I would agree that sampling from the NLLFS and simply comparing that to national averages does very little good and would not be worth including for the purposes of this discussion.

[Paprika]

That depends entirely on the shortcomings. Vaguely recalling Regnerus, the shortcomings have nothing to do with sample-size problems like the NLLFS

The point is that the same criticism "each has shortcomings making broad generalizations impossible" applies for both Regnerus' and the rest and those based on the NLLFS.

So long as the NLLFS and gold-standard studies all point in the same direction, there's nothing unethical about saying that the broad consensus supports the hypothesis

It's thoroughly unethical to level the same criticism - that they cannot be used to make broad generalisations - at opposing studies while whitewashing the same shortcomings in those that are cited to support the outcome you want.

[Tassman]

I've gotten different answers from Christians, but I assume you're of the opinion it's for children.

Of course it is, you know, like the Duggars.

[Paprika]

Yes, I would agree that sampling from the NLLFS and simply comparing that to national averages does very little good and would not be worth including for the purposes of this discussion.

Well then, the immediate question would be: "how many of the 40+ papers fall under this category and are therefore not worth inclusion?"

If the answer is "most", then the 'consensus' based on these older papers is complete bunkum. The few newer papers that are based on representative sampling have of course to be seriously engaged with. However I believe it clear the "clear consensus" based on an overwhelming number of studies across many decades was hardly established; without further substantiation you cannot portray the situation as 'oh, we have this existing consensus that later less-limited studies substantiate'.

[Starlight]

Paprika,

Any time a sociologist wanting to do a study tries to recruit people to be studied they're naturally going to get certain types of people signing up more often than others (eg people with enough spare time to participate, people wanting to show-off what great parents they are, people poor enough that the money paid to them for participating in the study is actually an incentive etc). But that's not really a problem if the control group is being recruited by the same type of snowball effect and in the same sort of areas (eg from the same preschools etc). So you might say "well only the good same sex parents will sign up for such studies" but it equally follows that only the good opposite-sex parents will sign up for such studies. As a result, sociologists have learned from their experiences of conducting these kinds of studies on other issues that the results do tend to be reliable.

A the particular strength of such studies is also that the sociologist can personally interact with the people and ask them a set of targeted questions and get a more personalized understanding of what the factors in play really are, whereas more impersonal studies (eg that analyse census data) can't necessarily draw conclusions from it if the crucial questions weren't asked in the original data set. It's also worth noting that snowball recruitment studies can sometimes get extremely large data sets - eg I saw the preliminary results of an ongoing Australian study on same-sex parenting, and they had recruited more than a quarter of all same-sex parents in Australia to participate in it, a rather impressive participation rate.

It is quite possible to imagine though that in any one particular study, the snowball effect might do something utterly unusual, and the study might be biased severely due to a systemic issue that ends up causing a whole lot of really good heterosexual parents to sign up and really terrible same sex parents to sign up. That issue is mitigated by having multiple such studies done by different researchers in different areas. If the outcomes are stable regardless of who obtained the sign-ups and how they did it, then that gives surety that the studies are all finding a real result. So if 40+ studies are done around the world by different researchers who all recruited using various different methods in various different countries, and yet the studies all found the same scientific result, then we can be very very sure that the result is actually real and not some strange phenomena resulting from the use of snowball sign-ups. For these reasons, the experts in the field are quite happy to take the cumulative results of such studies as compelling evidence, because this type of methodology is known to work perfectly fine in aggregate. The wonderful thing about the scientific method is repetition gives surety of reliability of the results.

Nonetheless it is of course nice to have some "representative" samples in order to check and confirm that the snowball effect hasn't done something totally bizarrely and outrageously unexpected, although we can be reasonable sure that multiple snowballed samples aren't really likely to be going wrong. The competently done nationally representative samples (of which I think there have been about four but I'd have to check) have all come back confirming no difference in outcomes for same-sex parenting, which has caused researchers in the field to largely shrug because they trusted the aggregate snowball samples anyway.

The incompetently done nationally representative samples have been the big issue as far as the media is concerned (though the sociologists have largely just dismissed them out of hand, because the error they make is so fundamental as to be laughable). The issue that has consistently dominated these incompetently done studies is family stability. One of the most thoroughly proven results in the history of the study of childhood development is that family instability (and by this I primarily mean divorce) negatively affects children. Children whose parents have divorced or separated do worse on average than children in stable households. That's a really basic finding which has been proven over and over again, and is really really well-known. What is also really really well-known is that in the present day, for a massive proportion of gay couples that are currently raising children together, those children were born in previous heterosexual relationships that subsequently failed. Thus heaps and heaps of gay couples are currently raising children that have lived through a prior heterosexual divorce. While adoption and IVF are on the rise as a way of gay couples getting children, and thus increasing numbers of gay parents are raising children who have not experienced a parental divorce, the proportion of gay couples currently raising children who have previously experienced divorce is still huge. So any study wanting to compare gay parenting to straight parenting absolutely needs to control for divorce - that's the number one thing that all sociologists know matters in childhood outcomes. There's absolutely no point comparing 1000 gay parents raising children who have experienced heterosexual divorces to 1000 straight parents raising children who have never experienced parental divorce... such a study is simply going to tell you that divorce has negative effects on the children. And that is pretty much precisely what the incompetently done nationally representative samples have done. The experts in the field simply respond with "?!?! Are we supposed to be taking your 'study' seriously?" If you're wanting to study straight parenting vs gay parenting you have to compare like with like: Either compare two groups of children who have experienced a divorce, or compare two groups of children who haven't. But Regnerus decided it would be a great idea to compare children of divorced parents (and mislabel them as children of gay parents) with children of non-divorced parents and then pretend that proved something about same-sex parenting. And Allen decided it would be a great idea to take census data of children being raised by same-sex parents (mostly who had experienced a previous heterosexual divorce) and compare it with data of children being raised by opposite-sex parents (mostly who hadn't experienced a previous heterosexual divorce), and imagined that he'd shown something about same-sex parenting... yet when other researchers took his dataset and controlled for family stability they found that his dataset showed divorce was bad and that there was no difference in same-sex vs opposite sex parenting.

That is why the experts in the field are prepared to issue definitive pronouncements on the subject. Their snowball samples all agree, and make a compelling case in aggregate. The representative samples agree too. And the 'studies' that disagree commit such a basic well-understood and fundamental error that there isn't really anything to discuss other than to roll one's eyes at them.

[Paprika]

So you might say "well only the good same sex parents will sign up for such studies" but it equally follows that only the good opposite-sex parents will sign up for such studies. As a result, sociologists have learned from their experiences of conducting these kinds of studies on other issues that the results do tend to be reliable.

The problem is that many studies have been done comparing snowball samples of homosexual parents to the national average parents, and these studies have been used to form the 'consensus'.

It's also worth noting that snowball recruitment studies can sometimes get extremely large data sets - eg I saw the preliminary results of an ongoing Australian study on same-sex parenting, and they had recruited more than a quarter of all same-sex parents in Australia to participate in it, a rather impressive participation rate.

Unfortunately most of the studies that form the 'consensus' have a sample size too small for any useful generalisations.

It is quite possible to imagine though that in any one particular study, the snowball effect might do something utterly unusual, and the study might be biased severely due to a systemic issue that ends up causing a whole lot of really good heterosexual parents to sign up and really terrible same sex parents to sign up. That issue is mitigated by having multiple such studies done by different researchers in different areas. If the outcomes are stable regardless of who obtained the sign-ups and how they did it, then that gives surety that the studies are all finding a real result.

A big if. So this demands of your side a close examination of the methodologies before any evidential weight can be placed on such studies using snowball samples.

Nonetheless it is of course nice to have some "representative" samples in order to check and confirm that the snowball effect hasn't done something totally bizarrely and outrageously unexpected, although we can be reasonable sure that multiple snowballed samples aren't really likely to be going wrong.

We cannot be reasonably sure because the snowball samples in general are likely to sample the more well-off homosexual families.

That is why the experts in the field are prepared to issue definitive pronouncements on the subject. Their snowball samples all agree, and make a compelling case in aggregate. The representative samples agree too. And the 'studies' that disagree commit such a basic well-understood and fundamental error that there isn't really anything to discuss other than to roll one's eyes at them.

The snowball samples mean very little since they're completely unrepresentative upon even surface inspection: of both sampling methods and sample size. Therefore the use of them to build a 'consensus' has been highly dishonest.

You're left with the representative samples to make your case. The ball is in yours (and Sam's court) to show that these studies are representative and don't have the limitations of the studies you criticise.

[Starlight]

http://factsaboutyouth.com/wp-conten...itself2010.pdf
http://www.soc.duke.edu/~jmoody77/20...uckner_ajs.pdf
http://www.sciencedirect.com/science...10440X06000952
http://journals.cambridge.org/action...2193200502674X
http://link.springer.com/article/10....A1010243318426
http://link.springer.com/article/10....508-008-9449-3

That's a somewhat random seeming list of links. I'm curious... what exactly in your mind are you claiming / suggesting / proving by posting these? I recall being a bit bemused last time you posted them also... 6 links to various different types of studies of a mix of qualities on a mix of topics... I'm not really sure what point you're trying to make or think you're making.

I'd remind you that the way science works is through repetition and that simply grabbing a single study at random and setting it up as the full and complete truth on any issue is not a good approach. So consider, for example, the question of whether divorce has negative effects on children who experience it or not. We might have 50 studies that have been conducted that conclude that it did effect the group they sampled. And then one study comes along and concludes that it didn't effect the group they sampled. Probably no one is then going to think "well we were all wrong about divorce all along, it's actually fine". Instead virtually everyone is going to think "well, something went a bit wonky with that last study. ~shrug~" Most scientists probably wouldn't have the slightest interest in even trying to work out what went wrong with that final study - everyone knows that in any data set there can be a few outliers, that sometimes a study just isn't done right or some error is made in the analysis of the data or wrong conclusions are drawn, or that sometimes a sample group has some sort of systemic bias that the researcher isn't aware of. That is why there are review articles, that compare the results of various different studies and compare their results and methodologies and try and draw some overall conclusions. Likewise any experts in the field are going to be aware of dozens upon dozens of studies.

So, on these issues in general, reading what the AMA, APA etc say in their court briefings is easily the best way to get an overall expert summary of what the cumulative results of all the different studies tells us (and the court submission is usefully targeted at a non-scientific audience and so makes easy reading once you're past the overly large tables of contents).

On most scientific issues, is it possible to find some study somewhere whose findings disagree with whatever the consensus on that issue happens to be? Of course. That's nearly always indicative that something's going wrong with that particular study. If a hundred people independently measure a distance and all conclude it's a mile, and one person measures it at two miles, then we can by and large safely assume that the one person somehow stuffed up their measurement and the question of how exactly they stuffed it up probably isn't a particular interesting one.

In general I'm not quite sure what you're anticipating I ought to conclude from your weblinks. Some of those studies look pretty badly conducted and I simply don't have any faith in the accuracy of their results. Others of them seem to repeat well-known results. Several of the authors appear to draw rather unlikely conclusions from their own data given there are some well-known results that would explain their data better. Etc. In short... weblinks to random studies on a variety of topics isn't any sort of meaningful scientific argument.

[JonathanL]

No. Your agreement is not needed because what changes is the zeitgeist. You might not agree with it at any particular moment but it will nevertheless drag you along behind it at some, acceptable to conservatives, distance. The old nutter fogies drop off the end of the tail to die; their death helping to drive the whole thing forwards like releasing the brakes on your moped …. to a position that will allow us to tackle the next thorny problem.

Are we there yet?

[Starlight]

The problem is that many studies have been done comparing snowball samples of homosexual parents to the national average parents, and these studies have been used to form the 'consensus'.

Is that actually the case, or are you merely speculating that it potentially could be? I would be extremely surprised if this were the case for any significant number of these studies. Especially given the relative ease of recruiting a heterosexual control group.

(Although I would note that if previous snowball samples of heterosexual parents had consistently and repeatedly found that they were in fact getting a nationally representative range of parenting quality represented and if this was a well-known finding within the field, then they could fairly safely compare the homosexual parenting outcomes with the national average. I personally have no clue whether this is the case or not.)

So this demands of your side a close examination of the methodologies before any evidential weight can be placed on such studies using snowball samples.

No, it merely relies on a having a number of such snowball samples. They're reliable in aggregate.

We cannot be reasonably sure because the snowball samples in general are likely to sample the more well-off homosexual families.

Are they? Is that known or are you speculating?

The snowball samples mean very little since they're completely unrepresentative upon even surface inspection: of both sampling methods and sample size.

I'm not sure what you mean when you refer to sample size here. Generally the snowball samples have dealt with larger numbers of gay parents than have the nationally representative samples. Because gay parents are relatively uncommon, taking a massive sample of the total population yields only a tiny tiny number of gay parents, so in order to get more gay parents in their samples the researchers instead preferred to go hunting for them via snowball recruitment techniques. This yields a higher number of total gay parents.

The ball is in yours (and Sam's court) to show that these studies are representative and don't have the limitations of the studies you criticise.

As I explained, the things that were wrong with the incompetently done studies was that they didn't control for divorce. The nationally representative studies that do control for divorce have found no difference between same sex and opposite sex parenting.

[Paprika]

Is that actually the case, or are you merely speculating that it potentially could be? I would be extremely surprised if this were the case for any significant number of these studies. Especially given the relative ease of recruiting a heterosexual control group.

I've quoted this before from Sam's study in this thread:

convenience or snowball samples are more common in the literature, and the most widely used data source is the National Longitudinal Lesbian Family Study (NLLFS). The NLLFS is based on interviews with donor-inseminated lesbian mothers five times from insemination or pregnancy to the child’s 17th birthday (e.g., Gartell and Bos 2010; Goldberg et al. 2011; van Gelderen et al. 2012a) and since 2002, 15 studies used these data. This recruitment strategy is considered acceptable given that few national surveys are large enough to include many children raised by same-sex parents. Relying on convenience samples means that the same-sex parents within these studies are not representative of all same-sex parents and represent only those who were targeted and agreed to participate, perhaps selective of the most highly functioning families. Yet, this approach does provide key insights into a group that is challenging to capture in large-scale surveys. At times, the findings from this sample are contrasted to results from a national sample of adolescents in the National Survey of Family Growth (NSFG) (Gartrell et al. 2012).

It falls to you and Sam to show that there are a significant number of studies where this does not occur.

No, it merely relies on a having a number of such snowball samples. They're reliable in aggregate.

Not necessarily. This isn't necessarily a case of random error where more measurements would reduce the overall error, but it could well be repeatedly encountering the systematic error in all the measurements (to use an analogy from the natural sciences).

Are they? Is that known or are you speculating?

See the box above: "Relying on convenience samples means that the same-sex parents within these studies are not representative of all same-sex parents and represent only those who were targeted and agreed to participate, perhaps selective of the most highly functioning families."

I'm not sure what you mean when you refer to sample size here. Generally the snowball samples have dealt with larger numbers of gay parents than have the nationally representative samples. Because gay parents are relatively uncommon, taking a massive sample of the total population yields only a tiny tiny number of gay parents, so in order to get more gay parents in their samples the researchers instead preferred to go hunting for them via snowball recruitment techniques. This yields a higher number of total gay parents.

Quite. Sam's link agrees with you, and so do I: representative sampling gets too few people, while snowball sampling is unrepresentative. Therefore (and here I am in agreement with the experts in Sam's link) there are great inherent methodological difficulties in investigating this subject, which is why I am very dubious (amongst other reasons) of the 'consensus': both of the validity of using the snowball samples and the representative samples to reach the standard conclusion amongst the progressives.

As I explained, the things that were wrong with the incompetently done studies was that they didn't control for divorce. The nationally representative studies that do control for divorce have found no difference between same sex and opposite sex parenting.

I have not relied on the 'incompetent' studies by Regenerus, Allen and the rest thus far in my argument. As to the nationally representative studies that you believe do not have the same or other major limitations, you are free to present them and we'll take a look.

[Adrift]

That's a somewhat random seeming list of links. I'm curious... what exactly in your mind are you claiming / suggesting / proving by posting these? I recall being a bit bemused last time you posted them also... 6 links to various different types of studies of a mix of qualities on a mix of topics... I'm not really sure what point you're trying to make or think you're making.

Sam said that he was unaware of any other studies, other than the Regnarus study, that demonstrated that there was developmental issues for children of same-sex partners. I was simply offering Sam links to some other studies that were not Regnarus.

The rest of your post has nothing to do with why I reposted these links in this thread for Sam, and it's largely a repeat of what you've already claimed in that other thread.

[Starlight]

I've quoted this before from Sam's study in this thread:

the most widely used data source is the National Longitudinal Lesbian Family Study (NLLFS). ...since 2002, 15 studies used these data. ....At times, the findings from this sample are contrasted to results from a national sample of adolescents in the National Survey of Family Growth (NSFG)

I'm not particularly worried by the passing comment that "At times" some of those studies use NSFG as a contrast/control group. There's no implication there that this is the only control group ever used. The idea that not a single author in 40+ studies bothered to use a sensible control group, when such control groups were fairly easy to obtain, is not really an idea I'm prepared to take very seriously because I know sociologists aren't universally massively incompetent and they do tend to understand basic science-101 concepts like control groups.

I am, however, slightly taken aback by the comment that 15 studies have been published on what is essentially the same data set (albeit it has evolved slightly over time as the children have aged). Obviously those 15 are not independent of each other, and should be considered a single longitudinal study.

See the box above: "Relying on convenience samples means that the same-sex parents within these studies are not representative of all same-sex parents and represent only those who were targeted and agreed to participate, perhaps selective of the most highly functioning families."

Right, so it's pure speculation. It might have selected for better parents, it might have selected for worse parents. Who knows?

there are great inherent methodological difficulties in investigating this subject, which is why I am very dubious (amongst other reasons) of the 'consensus'

Well that's really where the huge number of different studies done comes into it's own. Yes, any given study can be questioned. But when dozens of studies conducted in a dozen countries by different researchers using different methods all gave the same conclusion, then we can be sure of the results.

I'd note that another way of studying the topic is to study heterosexual parents (who have historically been studied a looooooooot) and examine how different gender roles among the parents affects childhood development. eg you could compare egalitarian parents to parents with stricter gender divisions etc. So if indeed there were truths about important parental gender roles that existed, sociologists who have studied heterosexual parenting ad nauseum over the 20th century really really ought to have found them, because ideas about gender roles and gender stereotypes have been of great political and social concern throughout the 20th century. As a result of such extensive and exhaustive studies of heterosexual parenting, sociologists seem to feel very certain in their conclusions about what does and doesn't matter. What definitely does matter is: family stability (lack of divorce); a loving relationships between the parents and between the parents and the child; and financial resources to provide for the child. But to the best of my knowledge, sociologists don't think that gender roles in parenting have any effect or importance whatsoever, and that is something they would definitely studied and looked at a lot. Knowing that the conformity or lack of it of heterosexual parents to gender roles or stereotypes has no effect on childhood outcomes, allows a pretty strong conclusion that same-sex parenting isn't going to cause any differences.

[Starlight]

Sam said that he was unaware of any other studies, other than the Regnarus study, that demonstrated that there was developmental issues for children of same-sex partners. I was simply offering Sam links to some other studies that were not Regnarus.

Only 3 of your 6 links appear at all relevant to that topic. The other 3 do not touch on children of same-sex partners (and instead put forth two, quite different, theories about what makes some people gay). Hence my confusion.

The three that do talk about children of same-sex partners, claim that such children are more likely to eventually identify as gay. If being gay has any genetic component at all, then such a finding would not be at all surprising - of course children of gay people would be more likely to be gay, just as children of tall people are more likely to be tall. However, overall, the evidence appears pretty mixed on the issue, and numerous studies have found no difference in the rate of children of same-sex partners identifying as gay. The overall consensus seems to be in favor of the view that gay people don't have gay children at any higher rate than straight people, as I mentioned last time.

[Paprika]

I'm not particularly worried by the passing comment that "At times" some of those studies use NSFG as a contrast/control group. There's no implication there that this is the only control group ever used. The idea that not a single author in 40+ studies bothered to use a sensible control group, when such control groups were fairly easy to obtain, is not really an idea I'm prepared to take very seriously because I know sociologists aren't universally massively incompetent and they do tend to understand basic science-101 concepts like control groups.

I'm afraid I don't share that same faith especially when strong ideological biases come into play.

I am, however, slightly taken aback by the comment that 15 studies have been published on what is essentially the same data set (albeit it has evolved slightly over time as the children have aged). Obviously those 15 are not independent of each other, and should be considered a single longitudinal study.

Well, they wanted numbers to make it seem like there was an overwhelming consensus. It's at least somewhat dishonest, I think you'll agree.

Right, so it's pure speculation. It might have selected for better parents, it might have selected for worse parents. Who knows?

It's very, very important because if the sampling is not representative then the conclusion that the experts want to draw from it likely does not follow. You can hardly brush it off with a 'who knows?'

Well that's really where the huge number of different studies done comes into it's own. Yes, any given study can be questioned. But when dozens of studies conducted in a dozen countries by different researchers using different methods all gave the same conclusion, then we can be sure of the results.

Above, you point out that 15 of the studies should count as only one; I do really doubt 'huge'. Furthermore using "different methods" in and of itself does not increase reliability because the methodologies of each study must be critically examined before evidentiary weight can be put on them. That is, quantity does not necessarily imply quality.

But to the best of my knowledge, sociologists don't think that gender roles in parenting have any effect or importance whatsoever, and that is something they would definitely studied and looked at a lot.

I look forward to you demonstrating that.