One thing I found I wished for and didn’t find, though, is a description of the underlying mechanics. You described what and why, but not how. Do you think that can be usefully expressed in a couple of paragraphs or it’s too complicated for that? The article already assumes a fair amount of background knowledge.
I’m not sure it can. I’ve read many different descriptions and looked at the math, but it’s a very different approach from the twin-based variance components estimation procedures I’ve managed to beat some understanding into my head of, and while I’ve worked with multilevel models & random effects in other contexts, the verbal descriptions of using multilevel models for estimating heritability just don’t make sense to me. (Judging from Visscher’s commentary paper, I may not be the only one having this problem.) I think my understanding of linear models and matrices may be too weak for it to click for me.
One problem I can see at first glance that the article doesn’t look like a Wikipedia article, but as a textbook or part of a publication. The goal of a Wikipedia article should be for a wide audience to understand the basics of something, and not a treatise only experts can comprehend.
What you wrote seems to be an impressive work, but it should be simplified (or at least the introduction of it), so that even non-experts can have a chance to at least learn what it is about.
1 full day. And I guess a few hours today checking edits other people made, tweaking parts of the article, responding to comments, etc. Plus, of course, all the background work that went into being able to write it in the first place… (‘How long, Mr Whistler?’) For example, I spent easily a week researching intelligence GCTAs and measurement error for my embryo selection cost-benefit analysis, which I could mostly copy-paste into that article. (I wanted an accurate GCTA estimate to put an upper bound on how much variance SNPs could ever explain and thus how much gain was possible per embryo. This required meta-analyzing GCTA estimates to get a stable point estimate and then correction for measurement error because a lot of the estimates are using imperfect measurements of intelligence.)
EDIT: and of course, after saying that, I then spent what must have been several other days working on digging up even more citations, improving related articles, and debating heritability and other stuff on Reddit...
It is my understanding that due to ethical concerns, the scientific field of psychology does not have a data collection methodology capable of distinguishing between effects caused by the parents’ genes and effects caused by the parents’ actions, and as such, no possible statistical approach will give a correct answer on the heritability of traits caused by the latter, like schizophrenia a.k.a. religion or intelligence. In order to clear up my “misunderstandings and ignorance”, you will need to demonstrate an approach that can, at the very least, successfully disprove genetic contribution in circumcision.
I think you need to read up a little more on behavioral genetics. To point out the obvious, besides adoption studies (you might benefit from learning to use Google Scholar) and and more recent variants like using sperm donors (a design I just learned about yesterday), your classic twin study design and most any ‘within-family’ design does control for parental actions, because they have the same parents. eg if a trait is solely due to parental actions, then monozygotic twins should have exactly the same concordance as dizygotic twins despite their very different genetic overlaps, because they’re born at the same time to the same parents and raised the same.
More importantly, the point of GCTA is that by using unrelated strangers, they are also affected by unrelated parents and unrelated environments. So I’m not sure what objection you seem to have in mind.
Sorry if I’m misunderstanding the method, but doesn’t it work something like finding strangers who have common genetics by chance?
If so, then 2 jews are more likely to have common genetics than chance, and also more likely to be circumcised. So it would appear that circumcision is genetic, when in fact it’s cultural.
It works by finding common genetics up to a limit of relatedness like fourth-cousin level. I think some Jewish groups may be sufficiently inbred/endogamous for long enough periods that it might not be possible to run GCTA with the usual cutoff since they’ll all be too related to each other. Population structure beyond that is dealt with by the usual approach of subtracting out 10 or 20 principal components and including them to control for that. This is a bit ad hoc but does work well in GWASes and gets rid of that problem, as indicated by the fact that the hits replicate within-family where the population structure is equalized by design and also have a good track record cross-racially/country too: https://www.reddit.com/r/science/comments/4kf881/largestever_genetics_study_shows_that_genetic/d3el0p2
It is my understanding that due to ethical concerns, the scientific field of psychology does not have a data collection methodology capable of distinguishing between effects caused by the parents’ genes and effects caused by the parents’ actions
Your understanding looks silly. It is rather obvious that not all children are brought up by their parents and that has been used in a number of studies. In fact, many classic identical-twins studies rely on being able to find genetically identical people who were brought up in different circumstances (including different parents).
Yes, it’s obvious. That’s why it was surprising when I couldn’t find a single study on schizophrenia where all children were separated from the parents immediately after birth. Feel free to enlighten me.
Misunderstandings and ignorance of GCTA seem to be quite pervasive, so I’ve tried to write a Wikipedia article on it: https://en.wikipedia.org/wiki/GCTA
Thanks for doing the frustrating work.
(The first and only comment so far is, more or less, “delete this article, because I don’t care”. Ugh.)
Yeah, that was weird. Almost as soon as I posted it, too. And the IP has only made 1 edit before, so it’s not some auto-troll.
Thank you.
One thing I found I wished for and didn’t find, though, is a description of the underlying mechanics. You described what and why, but not how. Do you think that can be usefully expressed in a couple of paragraphs or it’s too complicated for that? The article already assumes a fair amount of background knowledge.
I’m not sure it can. I’ve read many different descriptions and looked at the math, but it’s a very different approach from the twin-based variance components estimation procedures I’ve managed to beat some understanding into my head of, and while I’ve worked with multilevel models & random effects in other contexts, the verbal descriptions of using multilevel models for estimating heritability just don’t make sense to me. (Judging from Visscher’s commentary paper, I may not be the only one having this problem.) I think my understanding of linear models and matrices may be too weak for it to click for me.
One problem I can see at first glance that the article doesn’t look like a Wikipedia article, but as a textbook or part of a publication. The goal of a Wikipedia article should be for a wide audience to understand the basics of something, and not a treatise only experts can comprehend.
What you wrote seems to be an impressive work, but it should be simplified (or at least the introduction of it), so that even non-experts can have a chance to at least learn what it is about.
I don’t think this is true. Wikipedia is a collection of knowledge, not a set of introductory articles.
See e.g. the Wikipedia pages on intermediate-to-high statistical concepts and techniques, e.g. copulas).
Good god, how long did that take to write?
1 full day. And I guess a few hours today checking edits other people made, tweaking parts of the article, responding to comments, etc. Plus, of course, all the background work that went into being able to write it in the first place… (‘How long, Mr Whistler?’) For example, I spent easily a week researching intelligence GCTAs and measurement error for my embryo selection cost-benefit analysis, which I could mostly copy-paste into that article. (I wanted an accurate GCTA estimate to put an upper bound on how much variance SNPs could ever explain and thus how much gain was possible per embryo. This required meta-analyzing GCTA estimates to get a stable point estimate and then correction for measurement error because a lot of the estimates are using imperfect measurements of intelligence.)
EDIT: and of course, after saying that, I then spent what must have been several other days working on digging up even more citations, improving related articles, and debating heritability and other stuff on Reddit...
How does this reject the genetic factors causing circumcision in Jews?
What?
It is my understanding that due to ethical concerns, the scientific field of psychology does not have a data collection methodology capable of distinguishing between effects caused by the parents’ genes and effects caused by the parents’ actions, and as such, no possible statistical approach will give a correct answer on the heritability of traits caused by the latter, like schizophrenia a.k.a. religion or intelligence. In order to clear up my “misunderstandings and ignorance”, you will need to demonstrate an approach that can, at the very least, successfully disprove genetic contribution in circumcision.
I think you need to read up a little more on behavioral genetics. To point out the obvious, besides adoption studies (you might benefit from learning to use Google Scholar) and and more recent variants like using sperm donors (a design I just learned about yesterday), your classic twin study design and most any ‘within-family’ design does control for parental actions, because they have the same parents. eg if a trait is solely due to parental actions, then monozygotic twins should have exactly the same concordance as dizygotic twins despite their very different genetic overlaps, because they’re born at the same time to the same parents and raised the same.
More importantly, the point of GCTA is that by using unrelated strangers, they are also affected by unrelated parents and unrelated environments. So I’m not sure what objection you seem to have in mind.
Sorry if I’m misunderstanding the method, but doesn’t it work something like finding strangers who have common genetics by chance?
If so, then 2 jews are more likely to have common genetics than chance, and also more likely to be circumcised. So it would appear that circumcision is genetic, when in fact it’s cultural.
It works by finding common genetics up to a limit of relatedness like fourth-cousin level. I think some Jewish groups may be sufficiently inbred/endogamous for long enough periods that it might not be possible to run GCTA with the usual cutoff since they’ll all be too related to each other. Population structure beyond that is dealt with by the usual approach of subtracting out 10 or 20 principal components and including them to control for that. This is a bit ad hoc but does work well in GWASes and gets rid of that problem, as indicated by the fact that the hits replicate within-family where the population structure is equalized by design and also have a good track record cross-racially/country too: https://www.reddit.com/r/science/comments/4kf881/largestever_genetics_study_shows_that_genetic/d3el0p2
Your understanding looks silly. It is rather obvious that not all children are brought up by their parents and that has been used in a number of studies. In fact, many classic identical-twins studies rely on being able to find genetically identical people who were brought up in different circumstances (including different parents).
Yes, it’s obvious. That’s why it was surprising when I couldn’t find a single study on schizophrenia where all children were separated from the parents immediately after birth. Feel free to enlighten me.
Bzzzzz, I am sorry, you must have confused me with your research assistant. Please try again.