62 ms·

Unfortunately the field of observational statistics called "psychometrics" which has "proved" this using "twin studies" is largely bunk. The premise of those s

Unfortunately the field of observational statistics called "psychometrics" which has "proved" this using "twin studies" is largely bunk. The premise of those studies is linear additive effects from each causal factor, which is almost certainly false; and in general, there is no way of establishing causation without interventional experiments. If the effects were plausibly non-linear, we would need 1 trillion+ data points even to establish statistically significant differences. These fields are all cargo-cult stats, putting data into GUIs and printing pro-forma reports. The people involved are in way over their heads, to the degree that there isnt anything to "reproduce" let alone try to reproduce the experiments: they dont have any. The whole thing is little more than principle component analysis on astrological star data.

anonymouskimmer 3y ago
> The whole thing is little more than principle component analysis on astrological star data. When and where we are born definitely has an effect on personality, irrespective of family, in most cases. Astrological star data is a proxy for the when and where, and while thus would not be causative, would often be correlative. So it makes sense to do PCA on it as you'll identify a lot of good correlations.
- mjburgess 3y ago
  You are correct that star positions are proxies. The problem in the case of psychometrics is that the claim is that the stars (here: genes) arent proxies, but bonafide causes. So in the case of astrological analysis we would, if we wished to change population personalities change, say where people were born -- in the case of psychometrics we're told that we have no control over the causes whatsoever (genes). In the vast vast majority of observational genetic studies genes are just proxies, in the the same way the star signs are. In other cases, they're mere correlations. In the case of IQ (and basically all social genetic studies) it is highly likely that any correlated genes are merely proxies for geographic-historical distributions of peoples over time and it was the long-time-scale geographic and historical effects (culture, education, wealth, famine, poverty, war...) that we see in attainment. (Setting aside that IQ tests aren't valid above average IQ and have extremely bad properties both at the individual and population level for most of the measured range. )
  - anonymouskimmer 3y ago
    We do know that genes are a cause of IQ (we can see this comparing our nearest non-Homo relatives. But yes, we are far away from saying which genes do cause it, much less how they cause it, and how they interact with the environment and other genes to cause it.
    - mjburgess 3y ago
      IQ is a test, it isnt intelligence and isnt even a measure of intelligence. It is valid only for establishing certain sorts of mental retardation in a human population. To say that "genes cause IQ" is several levels of confusion. The claim has nothing to do with whether genes determine the physiological body structure, the claim is whether relevant aspects of human intelligence are a product of the at-birth or developmental structure -- or of that acquired from the environemnt, ie., neural plasticity / brain structure / etc. Almost all evidence we have is that relevant aspects of human intelligence are environmental, ie., acquired. There is little evidence of the alternative because the experiments needed to establish it are large impossible or unethical. So it might be true, but we have no evidence of it. Insofar as any claims are made about genes and IQ they amount to associating markers of geographical-historical migration (ie., breeding across time) with a very bad proxy measure of abstract reasoning which, at best, only detects deficits with any reliabiliy. So we have proxies on top of correlations on top of terrible measures of anything with zero reliability in the range of interest. This whole thing is operating nearly at fraudulent levels of "research". There's very very little this field can say scientifically, since there arent any experiments. It's all pseudoscientific uses of stats. To simplify matters, suppose that we know for sure that all above-average differences in IQ are 100% environmental/acquired-during-lifetime. Now, so what do twin studies show, on this assumption? They show that genes must be correlating with these environmental measures, and twin studies give us an 'effect size' for geographical-historical patterns of reporduction. So twin studies, even if they were valid (and their assumption of linearity renders them invalid), do not decide the matter. You need indpednent evidence of causal mechanisms from experiments, which doesnt exist.
      - sokoloff 3y ago
concordDance 3y ago
> The premise of those studies is linear additive effects from each causal factor I don't understand this. Why would you need linear additive effects to think that if identical twins are more similar than fraternal twins and half siblings are less similar than full siblings along some axis then there is probably a genetic component?
JumpCrisscross 3y ago
> Unfortunately the field of observational statistics called "psychometrics" which has "proved" this using "twin studies" is largely bunk Do you have a credible source for this claim? The criticism I'm seeing is largely constrained [1]. [1] https://en.wikipedia.org/wiki/Twin_study#Criticism https://en.wikipedia.org/wiki/Twin_study#Criticism

anonymouskimmer 3y ago

> The whole thing is little more than principle component analysis on astrological star data. When and where we are born definitely has an effect on personality, irrespective of family, in most cases. Astrological star data is a proxy for the when and where, and while thus would not be causative, would often be correlative. So it makes sense to do PCA on it as you'll identify a lot of good correlations.

mjburgess 3y ago
You are correct that star positions are proxies. The problem in the case of psychometrics is that the claim is that the stars (here: genes) arent proxies, but bonafide causes. So in the case of astrological analysis we would, if we wished to change population personalities change, say where people were born -- in the case of psychometrics we're told that we have no control over the causes whatsoever (genes). In the vast vast majority of observational genetic studies genes are just proxies, in the the same way the star signs are. In other cases, they're mere correlations. In the case of IQ (and basically all social genetic studies) it is highly likely that any correlated genes are merely proxies for geographic-historical distributions of peoples over time and it was the long-time-scale geographic and historical effects (culture, education, wealth, famine, poverty, war...) that we see in attainment. (Setting aside that IQ tests aren't valid above average IQ and have extremely bad properties both at the individual and population level for most of the measured range. )
- anonymouskimmer 3y ago
  We do know that genes are a cause of IQ (we can see this comparing our nearest non-Homo relatives. But yes, we are far away from saying which genes do cause it, much less how they cause it, and how they interact with the environment and other genes to cause it.
  - mjburgess 3y ago
    IQ is a test, it isnt intelligence and isnt even a measure of intelligence. It is valid only for establishing certain sorts of mental retardation in a human population. To say that "genes cause IQ" is several levels of confusion. The claim has nothing to do with whether genes determine the physiological body structure, the claim is whether relevant aspects of human intelligence are a product of the at-birth or developmental structure -- or of that acquired from the environemnt, ie., neural plasticity / brain structure / etc. Almost all evidence we have is that relevant aspects of human intelligence are environmental, ie., acquired. There is little evidence of the alternative because the experiments needed to establish it are large impossible or unethical. So it might be true, but we have no evidence of it. Insofar as any claims are made about genes and IQ they amount to associating markers of geographical-historical migration (ie., breeding across time) with a very bad proxy measure of abstract reasoning which, at best, only detects deficits with any reliabiliy. So we have proxies on top of correlations on top of terrible measures of anything with zero reliability in the range of interest. This whole thing is operating nearly at fraudulent levels of "research". There's very very little this field can say scientifically, since there arent any experiments. It's all pseudoscientific uses of stats. To simplify matters, suppose that we know for sure that all above-average differences in IQ are 100% environmental/acquired-during-lifetime. Now, so what do twin studies show, on this assumption? They show that genes must be correlating with these environmental measures, and twin studies give us an 'effect size' for geographical-historical patterns of reporduction. So twin studies, even if they were valid (and their assumption of linearity renders them invalid), do not decide the matter. You need indpednent evidence of causal mechanisms from experiments, which doesnt exist.
    - sokoloff 3y ago

concordDance 3y ago

> The premise of those studies is linear additive effects from each causal factor I don't understand this. Why would you need linear additive effects to think that if identical twins are more similar than fraternal twins and half siblings are less similar than full siblings along some axis then there is probably a genetic component?

JumpCrisscross 3y ago

> Unfortunately the field of observational statistics called "psychometrics" which has "proved" this using "twin studies" is largely bunk Do you have a credible source for this claim? The criticism I'm seeing is largely constrained [1]. [1] https://en.wikipedia.org/wiki/Twin_study#Criticism https://en.wikipedia.org/wiki/Twin_study#Criticism