Quantifying associations between socio-spatial factors and cognitive development in the ABCD cohort

Area of Science:

• Computational social science
• Developmental neuroscience
• Public health modeling

Background:

• Generative population models struggle with generalizability across diverse demographic groups.
• Existing models often fail to capture sociodemographic variability, limiting real-world applications.

Purpose of the Study:

• To propose a diversity-aware population modeling framework to improve generalizability for underrepresented groups.
• To guide targeted public health and education strategies by estimating subgroup-level effects.
• To capture sociodemographic variability in population models.

Main Methods:

• Leveraging Bayesian multilevel regression and post-stratification.
• Quantifying inter-individual differences in socioeconomic status and cognitive development relationships.
• Incorporating US Census data for enhanced subgroup interpretability in the Adolescent Brain Cognitive Development Study.

Main Results:

• Post-stratification improved model prediction interpretability for underrepresented groups.
• The framework ensured predictions were not skewed by overly heterogeneous or homogeneous subgroup representations.
• Bayesian multilevel modeling combined with post-stratification validated reliability and explained sociodemographic disparities.

Conclusions:

• A diversity-aware population modeling framework is crucial for equitable applications.
• Combining Bayesian multilevel modeling with post-stratification offers a robust approach to understanding sociodemographic disparities.
• This framework can enhance the reliability and holistic explanation of population-level effects across diverse groups.