Population neuroscience: Challenges and recommendations for researchers
Population Neuroscience is the interdisciplinary union of two seemingly distant fields -- basic neuroscience and population sciences, like demography -- that examine micro-level neural mechanisms in macro-level social contexts. Yet, Faculty Associate Arianna Gard argues that "a truly 'Population Neuroscience' approach would leverage the strengths of both fields to better understand how humans interact with and are shaped by the world. Studying the brain can reveal mechanisms of development, but appropriate attention to sampling and recruitment will ensure that this science is generalizable and impactful beyond our study samples."
Gard recently reviewed the state of this field in an article published in Developmental Cognitive Neuroscience, highlighting its shortcomings and offering suggestions to analyze population-level neuroimaging data effectively.
The dominance of convenience sampling in Psychology and Neuroscience has contributed to the under-representation of non-WEIRD (Western, Educated, Industrial, Rich, and Democratic) individuals. A popular avenue to counteract such discrepancies has been the advent of multi-site, population-based neuroimaging studies, like the Adolescent Brain and Cognitive Development℠ (ABCD) Study, that use a complex sampling design to capture a study sample that is more representative of the target population. The ABCD Study is one of the largest neuroimaging studies to date with data collected from nearly 12,000 youth across 21 geographically diverse sites in the United States.
Gard argues, however, that such studies may still capture sampling bias, and that this bias may be further embroiled by common exclusion criteria in neuroimaging analyses. Using the ABCD Study to test this claim, she identified that Hispanic or Latino/a youth, youth with a household income less than 200% of the federal poverty line, and youth of divorced or never-married caregivers were under-represented in the ABCD sample, compared to ACS-estimated proportions of same-age youth in the same year. Moreover, the under-represented groups were disproportionately more likely to be excluded from neuroimaging analyses due to "unusable" (i.e., typically high motion) data. Gard highlights the risks of these biases by empirically demonstrating the inaccurate estimation of brain-environment associations (e.g., brain activity and socioeconomic status) when survey weights are not used and multilevel structures are ignored.
The larger sample size and broader recruitment strategy of population-based neuroimaging studies offer a vital avenue for understanding brain-environment associations at the population-level. However, appropriately using these data requires analytic approaches that use survey weights and multilevel modeling to account for the statistical dependence of observations on nesting units, selection bias, non-response, and post-stratification.
Read the full article below:
