3 ms·

Using a PCA to prove an intrinsic point about ethnicity is as laughable as it gets. You're aware the vectors in a PCA are just linear combinations of the observ

Using a PCA to prove an intrinsic point about ethnicity is as laughable as it gets. You're aware the vectors in a PCA are just linear combinations of the observed features, right? You're just staring at the same data a different way. But of course I didn't expect any less from someone who quotes Slate Star Codex of all sources to prove a point.

rbecker 6y ago
> Using a PCA to prove an intrinsic point A what point? > You're aware the vectors in a PCA are just linear combinations of the observed features, right? Yes, I am. Do you have a reason to think that means PCA is bad at representing similarity and kinship because of this? As for the Slate Star Codex source - that wasn't to prove anything, just to define the "weak man" term. I would have used "straw man", but then you'd dig up some idiot that genuinely believes whatever it is you're debunking as proof that it's not a straw man.
- throrthaway 6y ago
  >A what point? https://www.merriam-webster.com/dictionary/intrinsic https://www.merriam-webster.com/dictionary/intrinsic >Do you have a reason to think that means PCA is bad at representing similarity and kinship because of this? I suggest you read up on the definition of "distance" and "similarity" before drinking the PCA kool-aid. You don't get to define an ad hoc distance just because it fits your ideas about ethnicity. But then, I only have the popgen community to back me up on this. What do you have? >As for the Slate Star Codex source - that wasn't to prove anything, just to define the "weak man" term. SSC, providing ammunition to online HBD proponents since 2013.
  - rbecker 6y ago
    There's nothing ad hoc about it - PCA is an extremely fundamental statistical tool, and commonly used in genetics, especially to evaluate population structure. A few random examples: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3392282/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3392282/ https://speciationgenomics.github.io/pca/ https://speciationgenomics.github.io/pca/ https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.0020190 https://journals.plos.org/plosgenetics/article?id=10.1371/jo... https://www.genetics.org/content/210/2/719 https://www.genetics.org/content/210/2/719 https://www.nature.com/articles/ng0508-491 https://www.nature.com/articles/ng0508-491 If that SSC reference bothers you so much, you can look at the Atlantic for a definition of weak man instead: https://www.theatlantic.com/daily-dish/archive/2009/07/the-weak-man-argument/199263/ https://www.theatlantic.com/daily-dish/archive/2009/07/the-w... And I know what intrinsic means, but I don't know what an "intrinsic point" is supposed to be, or what makes my point "intrinsic", as opposed to just a regular point.
    - throrthaway 6y ago
      Yes you don't need to explain my field to me, thank you very much. Are you going to post wikipedia links next? It still doesn't change the fact that PCA is extremely unrigorous for any analysis that goes beyond 'looking at pretty figures'. The vectors are literally and utterly meaningless. They're selected in an ad hoc way for each dataset. If you want to untangle population structure properly there are countless other more rigorous methods, although they are not without their flaws. But the fact that you want to use PC1 and PC2 plots to prove a weird point about ethnicity makes me think you either don't know what you're talking about and are just using the field as a front to advance your weird beliefs. Even Steve Hsu would be more rigorous than that

rbecker 6y ago

> Using a PCA to prove an intrinsic point A what point? > You're aware the vectors in a PCA are just linear combinations of the observed features, right? Yes, I am. Do you have a reason to think that means PCA is bad at representing similarity and kinship because of this? As for the Slate Star Codex source - that wasn't to prove anything, just to define the "weak man" term. I would have used "straw man", but then you'd dig up some idiot that genuinely believes whatever it is you're debunking as proof that it's not a straw man.

throrthaway 6y ago
>A what point? https://www.merriam-webster.com/dictionary/intrinsic https://www.merriam-webster.com/dictionary/intrinsic >Do you have a reason to think that means PCA is bad at representing similarity and kinship because of this? I suggest you read up on the definition of "distance" and "similarity" before drinking the PCA kool-aid. You don't get to define an ad hoc distance just because it fits your ideas about ethnicity. But then, I only have the popgen community to back me up on this. What do you have? >As for the Slate Star Codex source - that wasn't to prove anything, just to define the "weak man" term. SSC, providing ammunition to online HBD proponents since 2013.
- rbecker 6y ago
  There's nothing ad hoc about it - PCA is an extremely fundamental statistical tool, and commonly used in genetics, especially to evaluate population structure. A few random examples: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3392282/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3392282/ https://speciationgenomics.github.io/pca/ https://speciationgenomics.github.io/pca/ https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.0020190 https://journals.plos.org/plosgenetics/article?id=10.1371/jo... https://www.genetics.org/content/210/2/719 https://www.genetics.org/content/210/2/719 https://www.nature.com/articles/ng0508-491 https://www.nature.com/articles/ng0508-491 If that SSC reference bothers you so much, you can look at the Atlantic for a definition of weak man instead: https://www.theatlantic.com/daily-dish/archive/2009/07/the-weak-man-argument/199263/ https://www.theatlantic.com/daily-dish/archive/2009/07/the-w... And I know what intrinsic means, but I don't know what an "intrinsic point" is supposed to be, or what makes my point "intrinsic", as opposed to just a regular point.
  - throrthaway 6y ago
    Yes you don't need to explain my field to me, thank you very much. Are you going to post wikipedia links next? It still doesn't change the fact that PCA is extremely unrigorous for any analysis that goes beyond 'looking at pretty figures'. The vectors are literally and utterly meaningless. They're selected in an ad hoc way for each dataset. If you want to untangle population structure properly there are countless other more rigorous methods, although they are not without their flaws. But the fact that you want to use PC1 and PC2 plots to prove a weird point about ethnicity makes me think you either don't know what you're talking about and are just using the field as a front to advance your weird beliefs. Even Steve Hsu would be more rigorous than that