Re: Study: DNA test can predict whether you are gay
Posted: Tue Oct 13, 2015 7:38 pm
I notice that the BuzzFeed article gives the accuracy as 67%, not "nearly 70%." If, in fact, their "algorithm" were based on the sample of 47 pairs of guys, the results are even less defensible than I thought--from a purely mathematical point of view, ethics aside. However, the BuzzFeed article gives further information not in the posting at the start of the thread:
"They screened saliva samples collected years ago from male twin pairs. For 37 pairs, one of the twins was gay and the other straight, and for another 10 pairs, both twins were gay. These sorts of twin studies are useful in genetic research because twins generally share many of the same environmental influences and cultural upbringing."
In other words, the information I gave my statistician friend was misleading. The algorithm appears to have come first, then the test with the 47 pairs of twins, which is the proper way from the standpoint of mathematical propriety. If that's the case, then my friend says the results could be considered "statistically significant," though it would misclassify a large proportion of cases. At best, it's not dramatically better than flipping a coin. Also, to be accepted as science, the result needs to be independently repeated with a larger sample, as the BuzzFeed article notes:
"Several experts who were not involved in the research told BuzzFeed News that they were skeptical of the veracity of these results, particularly because the study was based on a relatively small sample of men.
“'All predictive models need replication with larger samples, and this one certainly does,' J. Michael Bailey, a professor of psychology at Northwestern University, told BuzzFeed News by email."
The researchers do appear to be sensitive to the ethical issues:
"Ngun, who is gay, agrees that this research is in its infancy, and says he has no intention of making a commercial test to predict sexuality. Still, he’s worried about the potentially ominous implications of his work — so much so, in fact, that he has decided to leave the field. He left academia last week, and plans to do research in industry that has nothing to do with the biology of sexual orientation.
“'It kind of, honestly, became a little bit troubling to me, what I was actually doing,' Ngun told BuzzFeed News. 'Having done this now, I could sort of foresee a not-so-happy outcome.'”
One can hope that the "algorithm" will never be published and will disappear into obscurity. Anybody who's skeptical that it would be abused needs to look at how many female fetuses are aborted in India, Georgia (the country), Pakistan, and China for the misfortune of having two X chromosomes. See https://en.wikipedia.org/wiki/Sex-selective_abortion
Sandi
"They screened saliva samples collected years ago from male twin pairs. For 37 pairs, one of the twins was gay and the other straight, and for another 10 pairs, both twins were gay. These sorts of twin studies are useful in genetic research because twins generally share many of the same environmental influences and cultural upbringing."
In other words, the information I gave my statistician friend was misleading. The algorithm appears to have come first, then the test with the 47 pairs of twins, which is the proper way from the standpoint of mathematical propriety. If that's the case, then my friend says the results could be considered "statistically significant," though it would misclassify a large proportion of cases. At best, it's not dramatically better than flipping a coin. Also, to be accepted as science, the result needs to be independently repeated with a larger sample, as the BuzzFeed article notes:
"Several experts who were not involved in the research told BuzzFeed News that they were skeptical of the veracity of these results, particularly because the study was based on a relatively small sample of men.
“'All predictive models need replication with larger samples, and this one certainly does,' J. Michael Bailey, a professor of psychology at Northwestern University, told BuzzFeed News by email."
The researchers do appear to be sensitive to the ethical issues:
"Ngun, who is gay, agrees that this research is in its infancy, and says he has no intention of making a commercial test to predict sexuality. Still, he’s worried about the potentially ominous implications of his work — so much so, in fact, that he has decided to leave the field. He left academia last week, and plans to do research in industry that has nothing to do with the biology of sexual orientation.
“'It kind of, honestly, became a little bit troubling to me, what I was actually doing,' Ngun told BuzzFeed News. 'Having done this now, I could sort of foresee a not-so-happy outcome.'”
One can hope that the "algorithm" will never be published and will disappear into obscurity. Anybody who's skeptical that it would be abused needs to look at how many female fetuses are aborted in India, Georgia (the country), Pakistan, and China for the misfortune of having two X chromosomes. See https://en.wikipedia.org/wiki/Sex-selective_abortion
Sandi