Talk:Genetic correlation

Correlation matrix?

It might be interesting to condense the very, very long list in a matrix format if it isn't too sparse. — Preceding unsigned comment added by 24.131.30.40 (talk) 03:07, 4 November 2016 (UTC)

I think it's too sparse. Most pairs are unrepresented, and in any case, would it be helpful to present a 120x120 matrix? --Gwern (contribs) 02:48 12 January 2017 (GMT)

Page is getting unwieldy

I'd like to start paring this page down and moving content to other articles, possibly. Objections? Many of the references are to primary source materials that are cited extensively, for example. Vrie0006 (talk) 18:35, 3 October 2018 (UTC)

I object. Nothing should be 'pared', as genetic correlations are an increasingly key part of modern quantitative genetics, having been unlocked by large-scale GWASes and summary statistics methods (as one can see simply by noting the distribution of dates). And there's nothing wrong with 'primary source materials', especially in a scientific context; in line with WP:PRIMARY, all the major interpretive information is thoroughly sourced to many review papers or textbooks, and the primary sources are cited for very specific, narrow claims (mostly just specific documented genetic correlations).

If you want to start picking out specific traits and shifting them over to their respective 'genetics of' pages, like Causes of schizophrenia#Genetics or Addiction#Genetic factors, that's fine but given how incomplete most such pages/sections are, it might be a better use of time to work on incorporating the basic twin/family/adoption/GWAS historical scientific background establishing genetic contributions before trying to merge in more detailed discussion of the genetic architecture. --Gwern (contribs) 22:04 3 October 2018 (GMT)

Hmmm, ok. This page is bloated. Primary source preprints are being used to support general statements. The list of observed genetic correlations I suggest should be its own page and be a useful sortable table. As it is, it is just not a useful list. At the very least these genetic correlations often have huge standard errors and many of these sources are primary. There are not reviews or replicated results. I'm just offering to help improve the page so that, in addition to all the minutiae currently here, lay-people can read and understand what genetic correlations are, and identify what the important review articles are. I'd humbly suggest the page is missing the forest for the trees at present. I say humbly b/c you seem to dominate this page and have strong opinions about it, and I don't want you to take offense. Has the page ever been reviewed for GA status, for example? That might be a helpful way to get an outside viewpoint on the current article. Vrie0006 (talk) 00:11, 4 October 2018 (UTC)

A fair number of the preprints have since been published but I don't update them since there's often not a fulltext and in any case, the published version is frequently worse (aside from being paywalled - page limits frequently result in brutal hackojbs of the preprint where large amounts of information are omitted, the methods sections are comically shrunk, the key contributions are obscured for lack of space for discussion). The standard errors may be large but they are decreasing over time and typically only the statistically-significant ones are reported. And there's lots of replications here just taking traits as literally reported (note how many pairs have many citations), and if you consider that many of these traits are only very slightly differently defined, the replication count is much larger. (How different are the 'years of education' and 'college attendance' traits, really? The latter is hardly anything but a noiser discretization of the former, and the genetic correlations reflect that, but they're still reported as different traits.) It is a technical topic, so there's only so much tat can be done, but I've tried to make it comprehensible, and the textbook chapters are referenced in the introduction. --Gwern (contribs) 22:49 6 October 2018 (GMT)

I would suggest preprints not be used at all and I disagree that preprints are better. The biggest problem with the page is the list. Why not have a table with three columns, trait 1, trait 2, and correlation + reference? (Maybe a fourth column of "domain" of the phenotype(s). Then put that on a separate page called "List of Reported Genetic Correlations" or something like that? I can't imagine anyone actually uses the list here in any meaningful way. Vrie0006 (talk) 02:18, 7 October 2018 (UTC)

I use the list routinely; just today I was looking for genetic correlations of longevity/intelligence. I know some researchers check it because they've told me. And there's a decent amount of page traffic and I doubt they're all interested in how one would use a rg to boost a GWAS...

I want to second that I use this page for reference when looking something up and know that many colleagues do the same. Deleet (talk) 18:49, 4 December 2018 (UTC)

I didn't start with a table because wikitables have always confused me and I didn't need the structure - I wasn't intending to include the actual correlation because it would be an unfathomable amount of work to dig numbers out of all the references. If all you have is trait 1/trait2<ref>, you don't need a table. But if you're volunteering... --Gwern (contribs) 02:39 10 October 2018 (GMT)