3 ms·

How did they discover that, given the curse of dimensionality? If the genetic contribution is significant, how do they know it's not some unreproducible model t

How did they discover that, given the curse of dimensionality? If the genetic contribution is significant, how do they know it's not some unreproducible model that is significant due to it having too many parameters (thousands of genes)? Or do they have a theoretical model of homosexuality and took just the genes that the model predicts should point to it?

buboard 7y ago
statistics
- qzx_pierri 7y ago
  [citation needed]
  - acdha 7y ago
    It’s a scientific study so the citation is literally present in the story if you want to learn what they did: http://dx.doi.org/10.1126/science.aat7693 http://dx.doi.org/10.1126/science.aat7693
- superpermutat0r 7y ago
  Curse of dimensionality and statistical significance are terms I used in my comment. I'm quite aware of the statistical methods and am a bit confused about how they avoid the dimensionality problem. I'm pretty sure I could get statistical significance for any trait with a highly dimensional dataset.
  - dooglius 7y ago
    Rather than making armchair dismissals, consider looking at the paper and its citations which discuss the statistical methods used. From a cursory glance, it looks like they use a standard model of how genes affect traits, which uses O(N) parameters for N genes, so the "curse of dimensionality" would not apply.
    - superpermutat0r 7y ago
      When there's thousands of genes then it applies.
vixen99 7y ago
Maybe akin to the IQ story. Polygenic scores have been shown in at least 20 studies to predict IQ & educational attainment. "Childhood intelligence is heritable, highly polygenic and associated with FNBP1L" https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3935975/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3935975/
bayesian_horse 7y ago
Inheritability of a trait can be inferred through statistic means, though I feel like there is too much confidence around such computations, given that it is virtually impossible to isolate environment from genetics in Humans. But when looking for specific mutations, the sum of all these effects is often just a small fraction of that inheritability, suggesting that the interaction of the genes and the environment is more complex than can be captured with such models.

buboard 7y ago

statistics

qzx_pierri 7y ago
[citation needed]
- acdha 7y ago
  It’s a scientific study so the citation is literally present in the story if you want to learn what they did: http://dx.doi.org/10.1126/science.aat7693 http://dx.doi.org/10.1126/science.aat7693
superpermutat0r 7y ago
Curse of dimensionality and statistical significance are terms I used in my comment. I'm quite aware of the statistical methods and am a bit confused about how they avoid the dimensionality problem. I'm pretty sure I could get statistical significance for any trait with a highly dimensional dataset.
- dooglius 7y ago
  Rather than making armchair dismissals, consider looking at the paper and its citations which discuss the statistical methods used. From a cursory glance, it looks like they use a standard model of how genes affect traits, which uses O(N) parameters for N genes, so the "curse of dimensionality" would not apply.
  - superpermutat0r 7y ago
    When there's thousands of genes then it applies.

vixen99 7y ago

Maybe akin to the IQ story. Polygenic scores have been shown in at least 20 studies to predict IQ & educational attainment. "Childhood intelligence is heritable, highly polygenic and associated with FNBP1L" https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3935975/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3935975/

bayesian_horse 7y ago

Inheritability of a trait can be inferred through statistic means, though I feel like there is too much confidence around such computations, given that it is virtually impossible to isolate environment from genetics in Humans. But when looking for specific mutations, the sum of all these effects is often just a small fraction of that inheritability, suggesting that the interaction of the genes and the environment is more complex than can be captured with such models.