Yes, the gay uncle hypothesis, that the gay phenotype has positive inclusive fitness, is absurd. But I think Elo was referring to the sexual antagonism hypothesis, that the gene increases fitness in women who carry it and decreases fitness in men. Then we are out of the realm of inclusive fitness and the tradeoff is 1:1. Moreover, if the female variant has 100% penetrance and the male variant only 1⁄3 penetrance, then there is a 3:1 advantage. So if gay men are down a child, female carriers only need to have 1⁄3 of an extra child.
In one sense 1⁄3 of an extra child is small. It would be pretty hard to measure that sisters of gay men have 1⁄6 of an extra child. The claim isn’t immediately inconsistent with the two observations of (1) gay equilibrium; and (2) not particularly fecund relatives. But in another sense, 1⁄3 of an extra child in reach is a huge fitness boost and I would expect natural selection to find another route to this opportunity. So I don’t believe this, either, but it’s a lot better than the gay uncle hypothesis.
Well, maybe. I don’t much like that either since you would expect some sex-linked adaptations to neutralize it in males on the Y chromosome (and you would expect to see low genetic correlation between female homosexuality and male homosexuality if they’re entirely different things) and I don’t think I’ve seen many plausible examples of sexual antagonism in humans. (Actually, none come to mind.) It’s better than overall inclusive fitness but still straining credulity for such a pervasive fitness-penalizing phenomenon.
In one sense 1⁄3 of an extra child is small. It would be pretty hard to measure that sisters of gay men have 1⁄6 of an extra child.
Why is that? 0.33 kids vs a mean of 2.1 and an SD of I dunno, 3 (lots of people have 0 kids, a fair number have 4-6), implies a fairly observable effect size with a sample requirement of n=1300 (power.t.test(power=0.8, d=0.33/3)), even less if you take advantage of within-family comparisons to control away some of that variance, and it doesn’t require a particularly exotic survey dataset—most studies which ask in as much detail as sexual orientation will also collect basic stuff like ‘number of offspring’.
Your power calculation was for an effect size of 1⁄3, which only makes sense if you know exactly which women have the gene.
The obvious test is to look at the children of sisters of gay vs straight men. But then the relationship is 1⁄2, so this cuts down the necessary advantage to 1⁄6. This has been done, but it is not at all standard, and I the sample size was too small. Aunts have the advantage of having being older and thus more likely to have completed their fertility and the disadvantage of being less related.
The fertility of mothers is more commonly measured: the number of siblings. And, indeed, it is often claimed that gay men come from large families, or at least that they have older brothers. But we still don’t know that the mother has the gene. As a warm-up, consider the case of penetrance still 1⁄3, but gays having no children, requiring female carriers to have an extra 2⁄3 to compensate. Then male carriers have 4⁄3 children and female carriers 8⁄3, so the gay man is 2⁄3 likely to have gotten the gene from his mother, 1⁄3 from his straight father. So his expected family size is an extra 4⁄9, which is measurable in a large sample.
Back to my model where gay men have 1 child and female carriers 7⁄3. Then the gay son is 7⁄12 likely to have gotten the gene from his mother, 4⁄12 from his straight father and 1⁄12 from his gay father. So a naive calculation says that his family size should be boosted by 1/3*7/12 and reduced by 1*1/12, thus net increased by 1⁄9, which is very small. (This is naive because the fertility of a gay man conditioning on his having a child probably depends on the exact distribution of fertility. This problem also comes up conditioning on the mother having a child, but with a small fertility advantage it probably doesn’t matter much comparing the mother of a gay and a straight.)
Your power calculation was for an effect size of 1⁄3, which only makes sense if you know exactly which women
Not sure I follow. I wasn’t talking about the gene, I was talking about the net fertility impact which must exist in the sisters to offset their gay brother’s lack of fitness. If you want to see if it exists, all you have to do is compare the sisters of gay men to non-gay men; either the former have enough extra babies to make up for it or not.
have the gene.
There is no single gene for homosexuality otherwise the pedigree studies would look much clearer than they do and not like a liability-threshold sort of thing, the linkage studies would’ve likely already found it, or 23andMe’s GWAS would’ve (homosexuality is so common that a single variant would have to be very common; they probably ran a GCTA since I know they’ve GCTAed at least 100 traits they haven’t published on but not whether they checked the correlation with chromosome length to check for polygenicity). So I’m not sure your calculations there are relevant.
Yes, of course all my calculations are under the simplifying assumption of a single gene. But under that assumption, sisters of gay men have only 1⁄2 chance of having the gene and so their expected additional number of babies is only 1⁄6. If you don’t think that this assumption is appropriate, you can suggest some other model and do a calculation. One thing I can guarantee you is that it won’t produce the number 1⁄3.
If you don’t think that this assumption is appropriate, you can suggest some other model and do a calculation. One thing I can guarantee you is that it won’t produce the number 1⁄3.
I think you’re missing the point. The effect size of the gene or genes is irrelevant, as is the architecture. There can be any distribution as long as there’s enough to be consistent with current genetic research on homosexuality having turned up few or no hits (linkage, 23andMe’s GWAS & GCTA, etc). The important question is merely: do their sisters have enough kids to via inclusive fitness make up for their own lack of kids? If the answer is no, you’re done with the sexual antagonism theory, so you only need to detect that. This is set by the fitness penalty of being homosexual, not by any multiplications. So if homosexuals have 1 fewer kid, then you need to detect 2 kids among their sisters, and so on. From that you do the power calculation.
So if homosexuals have 1 fewer kid, then you need to detect 2 kids among their sisters, and so on. From that you do the power calculation.
Back when you did the power calculation for 1⁄3 rather than 2, you didn’t believe that. This number 2 is wrong for three reasons:
Inclusive fitness is irrelevant to the antagonistic selection hypothesis. (factor of 2)
It ignores penetrance, which is clearly not 100%; it doesn’t matter how many children homosexuals have, but rather how many children (male) carriers of the gene(s) have. (factor of 3)
It ignores the fact that siblings only are only 1⁄2 related. The relevant gene(s) should only elevate the fertility of carriers, not all sisters. (factor of 2)
Yes, the gay uncle hypothesis, that the gay phenotype has positive inclusive fitness, is absurd.
But I think Elo was referring to the sexual antagonism hypothesis, that the gene increases fitness in women who carry it and decreases fitness in men. Then we are out of the realm of inclusive fitness and the tradeoff is 1:1. Moreover, if the female variant has 100% penetrance and the male variant only 1⁄3 penetrance, then there is a 3:1 advantage. So if gay men are down a child, female carriers only need to have 1⁄3 of an extra child.
In one sense 1⁄3 of an extra child is small. It would be pretty hard to measure that sisters of gay men have 1⁄6 of an extra child. The claim isn’t immediately inconsistent with the two observations of (1) gay equilibrium; and (2) not particularly fecund relatives. But in another sense, 1⁄3 of an extra child in reach is a huge fitness boost and I would expect natural selection to find another route to this opportunity. So I don’t believe this, either, but it’s a lot better than the gay uncle hypothesis.
Well, maybe. I don’t much like that either since you would expect some sex-linked adaptations to neutralize it in males on the Y chromosome (and you would expect to see low genetic correlation between female homosexuality and male homosexuality if they’re entirely different things) and I don’t think I’ve seen many plausible examples of sexual antagonism in humans. (Actually, none come to mind.) It’s better than overall inclusive fitness but still straining credulity for such a pervasive fitness-penalizing phenomenon.
Why is that? 0.33 kids vs a mean of 2.1 and an SD of I dunno, 3 (lots of people have 0 kids, a fair number have 4-6), implies a fairly observable effect size with a sample requirement of n=1300 (
power.t.test(power=0.8, d=0.33/3)), even less if you take advantage of within-family comparisons to control away some of that variance, and it doesn’t require a particularly exotic survey dataset—most studies which ask in as much detail as sexual orientation will also collect basic stuff like ‘number of offspring’.Your power calculation was for an effect size of 1⁄3, which only makes sense if you know exactly which women have the gene.
The obvious test is to look at the children of sisters of gay vs straight men. But then the relationship is 1⁄2, so this cuts down the necessary advantage to 1⁄6. This has been done, but it is not at all standard, and I the sample size was too small. Aunts have the advantage of having being older and thus more likely to have completed their fertility and the disadvantage of being less related.
The fertility of mothers is more commonly measured: the number of siblings. And, indeed, it is often claimed that gay men come from large families, or at least that they have older brothers. But we still don’t know that the mother has the gene. As a warm-up, consider the case of penetrance still 1⁄3, but gays having no children, requiring female carriers to have an extra 2⁄3 to compensate. Then male carriers have 4⁄3 children and female carriers 8⁄3, so the gay man is 2⁄3 likely to have gotten the gene from his mother, 1⁄3 from his straight father. So his expected family size is an extra 4⁄9, which is measurable in a large sample.
Back to my model where gay men have 1 child and female carriers 7⁄3. Then the gay son is 7⁄12 likely to have gotten the gene from his mother, 4⁄12 from his straight father and 1⁄12 from his gay father. So a naive calculation says that his family size should be boosted by 1/3*7/12 and reduced by 1*1/12, thus net increased by 1⁄9, which is very small. (This is naive because the fertility of a gay man conditioning on his having a child probably depends on the exact distribution of fertility. This problem also comes up conditioning on the mother having a child, but with a small fertility advantage it probably doesn’t matter much comparing the mother of a gay and a straight.)
Not sure I follow. I wasn’t talking about the gene, I was talking about the net fertility impact which must exist in the sisters to offset their gay brother’s lack of fitness. If you want to see if it exists, all you have to do is compare the sisters of gay men to non-gay men; either the former have enough extra babies to make up for it or not.
There is no single gene for homosexuality otherwise the pedigree studies would look much clearer than they do and not like a liability-threshold sort of thing, the linkage studies would’ve likely already found it, or 23andMe’s GWAS would’ve (homosexuality is so common that a single variant would have to be very common; they probably ran a GCTA since I know they’ve GCTAed at least 100 traits they haven’t published on but not whether they checked the correlation with chromosome length to check for polygenicity). So I’m not sure your calculations there are relevant.
Yes, of course all my calculations are under the simplifying assumption of a single gene. But under that assumption, sisters of gay men have only 1⁄2 chance of having the gene and so their expected additional number of babies is only 1⁄6. If you don’t think that this assumption is appropriate, you can suggest some other model and do a calculation. One thing I can guarantee you is that it won’t produce the number 1⁄3.
I think you’re missing the point. The effect size of the gene or genes is irrelevant, as is the architecture. There can be any distribution as long as there’s enough to be consistent with current genetic research on homosexuality having turned up few or no hits (linkage, 23andMe’s GWAS & GCTA, etc). The important question is merely: do their sisters have enough kids to via inclusive fitness make up for their own lack of kids? If the answer is no, you’re done with the sexual antagonism theory, so you only need to detect that. This is set by the fitness penalty of being homosexual, not by any multiplications. So if homosexuals have 1 fewer kid, then you need to detect 2 kids among their sisters, and so on. From that you do the power calculation.
Back when you did the power calculation for 1⁄3 rather than 2, you didn’t believe that. This number 2 is wrong for three reasons:
Inclusive fitness is irrelevant to the antagonistic selection hypothesis. (factor of 2)
It ignores penetrance, which is clearly not 100%; it doesn’t matter how many children homosexuals have, but rather how many children (male) carriers of the gene(s) have. (factor of 3)
It ignores the fact that siblings only are only 1⁄2 related. The relevant gene(s) should only elevate the fertility of carriers, not all sisters. (factor of 2)