nostalgebraist

anyway, my basic beef w/ g is: i think a lot can go wrong when you do stats w/o an underlying causal model

basically i just agree with borsboom's "attack of the psychometricians" http://www.stat.cmu.edu/~brian/Pmka-Attack-V71-N3/pmka-2006-71.3-425-440-borsboom.pdf …

and i thought fitness/"f" was a clear example where we could do stats w/o modeling and not really learn much

in the book, you mention thomson then say g "exists" in any event; i think existence claims require models

unless it's the case that any factor i get out of PCA etc "exists," giving us a weird ontology full of...

...things with no clear tie to the casual structure we're more sure of (from physics etc)

I think we actually agree in every respect. By "exists" I just meant that the factor is there (it could've been otherwise, cf. Thurstone)...

...in the data, explains a lot of the variance (again, could've been otherwise), and predicts stuff IRL.

oh, you're here, ok. Just thought I could add thishttps://twitter.com/SilverVVulpes/status/846541663040364544 …

and the philosophical traditions of Alternative Earth that fitness factors are independent

and several popsci books arguing that,against evidence. With impact on gymns and athletes.

otherwise I think Stu has said the rest (including interest that g is 50% and not, say, 8%)

i guess i disagree w author of paper you quoted. "we have multiple faculties" != "statistical independence on tests"

e.g. in fitness example, "faculty physiology" is /true/ (muscles and heart are distinct etc) yet we see pos correlations