2 of 14

Background and Motivation

Aggregate-level studies find fairly robust and strong relationships between bio-geographical ancestry (BGA) and S/cognitive-related outcomes; European ancestry is associated with better outcomes as compared to Amerindian/African ancestry (e.g., Fuerst and Kirkegaard 2016a).
Across the Americas, self-identified race/ethnicity (SIRE) is similarly associated with outcomes (Lynn, 2008; Fuerst and Kirkegaard, 2016b); ‘White/European’ SIRE is positively correlated with outcomes relative to Amerindian and African ancestry.
Older admixture studies using phenotypic proxies for admixture confirm the pattern (Shuey, 1966; Malloy and Fuerst, under preparation).

3 of 14

Background and Motivation

New medical studies apply admixture analysis/mapping to investigate the association between medical/other outcomes and BGA.
BGA indexed with ancestrally informative markers (AIMs).
Narrative reports (e.g., González Burchard et al., 2005) note that individual level BGA is associated with S indicators.
S represents a potential environmental confound (though controlling for it also potentially invokes the sociologist's fallacy).
No one has reviewed/meta-analyzed these studies.

4 of 14

Proposed causal model

5 of 14

Method

Locate studies that include measures of global ancestry and S indicators.
Code these studies by their features.
Contact authors for papers that did not report the necessary statistics.
Exclude studies e.g., with low admixture (mean ancestry < 5%).
Meta-analyze the studies.

6 of 14

Included studies

Complex hierarchical data:

One study may have multiple samples
One sample may have multiple outcomes
One sample may have multiple ancestries

N studies = 41.
Total number of tested associations: N = 59,663, k = 64
Studies often do not report effect sizes. Failure to use methods that allow for the accumulation of scientific knowledge (Schmidt 1996, Curran 2009). Failure to share data, even upon request, make the error uncorrectable.

7 of 14

Directional meta-analyses

Aggregate directions within sample by coding positive as 1, negative as -1 and null as 0.
Results:

8 of 14

Correlational meta-analyses

Aggregated effect size within sample using medians.
N for correlational analyses:

N for European = 34,266.5
N for African = 28,533.5
N for Amerindian = 20,544.5

Simple results:

Random effects meta-analysis is superior.

9 of 14

Correlational meta-analysis: European ancestry

10 of 14

Correlational meta-analysis: Amerindian ancestry

11 of 14

Correlational meta-analysis: African ancestry

12 of 14

Moderators and statistical artefacts

Very high heterogeneity: I2: 92%
Too few studies for formal moderator analyses, including publication bias, but many options:

Different trait levels for same ancestry in different locations. E.g. Iberian European vs. Northern European.
Different cultural practices by country: race discrimination (e.g. affirmative action, colorism).

Reasons to suspect statistical artifacts with negative bias:

Many studies used relatively few AIMs to estimate ancestry: median N AIM = 106. Fewer AIMs generate less reliable estimates of ‘true’ ancestry (e.g., r(Ancestry 15 AIMs x Ancestry 50,000 AIMs) = 0.60 (Scharf et al., 2013)).
Outcomes measured using ordinals with few levels: median N levels = 6.
Variation/range restriction in ancestries.

Collinearity between ancestries, need multiple regression results or raw data.

13 of 14

Future studies

Results are causally ambiguous: differences due to phenotypic based discrimination, genes, or familial cultural transmission?
Can study colorism with BF/WF sibling studies (Jensen 1980); preliminary results: strong version does not hold (Fuerst and Kirkegaard 2016c).
Admixture mapping: look at the regions of the genome where the BGA x SES/CA association is the strongest e.g., Zou et al. (2015) found that ancestry related assortative mating in P. R. was the strongest in regions that coded for facial morphology.
Alternatively, can look at the BGA x outcome associations among siblings reared together (Malloy 2013).

14 of 14

References

Curran, Patrick J. (2009). Special issue of Psychological Methods: Multi-Study Methods for Building a Cumulative Psychological Science. Psychological Methods, 14(2).
Fuerst and Kirkegaard (2016a). Admixture in the Americas: Regional and National Differences. Mankind Quarterly, 56(3).
Fuerst and Kirkegaard (2016b). The genealogy of differences in the Americas. Mankind Quarterly, 56(3).
Fuerst and Kirkegaard (in preparation). Does skin brightness-based discrimination explain group differences in cognitive ability and educational attainment? A study of US siblings.
Frudakis, T. N., & Shriver, M. D. (2003). Compositions and methods for inferring ancestry. U.S. Patent and Trademark Office Publication Number 20040229231.
González Burchard, E., Borrell, L. N., Choudhry, S., Naqvi, M., Tsai, H. J., Rodriguez-Santana, J. R., ... & Arena, J. F. (2005). Latino populations: a unique opportunity for the study of race, genetics, and social environment in epidemiological research. American Journal of Public Health, 95(12), 2161-2168.
Jensen, Arthur, R. (1980). Uses of sibling data in educational and psychological research. American Educational Research Journal, 17(2), 153—170.
Lynn, R. (2008). The global bell curve: Race, IQ, and inequality worldwide. Washington Summit Publishers.
Malloy, Jason (2013). Cryptic Admixture, Mixed-Race Siblings, & Social Outcomes. Human Varieties.
Scharf, J. M., Yu, D., Mathews, C. A., Neale, B. M., Stewart, S. E., Fagerness, J. A., ... Pauls, D. L. (2013).Genome-wide association study of Tourette Syndrome. Molecular Psychiatry, 18 (6), 721–728.
Schmidt, Frank (1996). Statistical Significance Testing and Cumulative Knowledge in Psychology: Implications for Training of Researchers. Psychological Methods, 1(2).
Zou, J. Y., Park, D. S., Burchard, E. G., Torgerson, D. G., Pino-Yanes, M., Song, Y. S., … & Zaitlen, N. (2015). Genetic and socioeconomic study of mate choice in Latinos reveals novel assortment patterns. Proceedings of the National Academy of Sciences, 201501741.