2023-06-30
General intelligence (i.e., g-factor) was first proposed by Spearman (1904) to account for the “positive manifold” phenomenon of cognitive tests.
Given the term “intelligence” is hard to define and easily leads to confusion, many researchers (see Jensen (1998)) focused on the g (i.e., general intelligence) (Barbey, 2021; Haier, 2017).
Current cognitive scientists also defined intelligence as general cognitive ability (Barbey, 2021), which can be derived by multiple cognitive tasks
Two major influences: sampling of participants and sampling of cognitive tasks
Here we focus on the issue of tasks sampling
Factor analysis method
Task sampling representativeness
There are several different factor analysis modeling method:
Spearman Model
Bifactor Model
Orthogonalized Hierarchical Model
Jensen (1998) found that the g-factor scores is extremely stable among different methods, with correlation coefficients ranging from 0.991 to 1.000. Here we just focus the classical Spearman model for its simplicity.
Project | Number of Cognitive Tasks | Ref |
---|---|---|
UK Biobank (2004) | 4 | Cox (2019) |
HCP (2009) | 12 | Dubois (2018) |
Aging Brain Cohort (2021) | 5 (part of NIH toolbox) | Newman-Norlund (2021) |
ABCD Study (2015) | 10 (7 from NIH toolbox) | Thompson (2019) |
Behavior sample: 1730 participants (Mean age = 20.8, SD = 2.1, range: [16.8, 30.83]; Sex: 58.9% females, 41.1% males, 0.0% other)
FMRI sample: 731 participants (Mean age = 20.9, SD = 2.2, range: [17, 29]; Sex: 62.5% females, 37.5% males, 0.0% other)
The following is to test whether the correlation between the estimated g-factor scores and the brain functional connectivity can be improved by eliminating certain observed variables, e.g., those with the least factor loading.