Biogeographic Ancestry and Socioeconomic Outcomes in the Americas: A Meta-Analysis

Abstract

Narrative reports suggest that socioeconomic status (SES) is associated with biogeographic ancestry (BGA) in the Americas. If so, SES potentially acts as a confound that needs to be taken into account when evaluating the relation between medical outcomes and BGA. To explore how systematic BGA-SES associations are, a meta-analysis of American studies was conducted. 40 studies were identified, yielding a total of 64 independent samples with directions of associations, including 48 independent samples with effect sizes. An analysis of association directions found a high degree of consistency. The square root n-weighted directions were 0.83 (K = 36),-0.81 (K = 41) and-0.82 (K = 39) for European, Amerindian and African BGA, respectively. An analysis of effect size magnitudes found that European BGA was positively associated with SES, with a meta-analytic effect size of r = .18 [95% CI: .13 to .24, K = 28, n = 35,476.5], while both Amerindian and African BGA were negatively associated with SES, having meta-analytic effect sizes of-.14 [-.18 to-.10, K = 31, n = 28,937.5] and-.11 [-.15 to-.07, K = 28, n = 32,710.5], respectively. There was considerable cross-sample variation in effect sizes (mean I2 = 92%), but the sample size was not enough for performing credible moderator analysis. Implications for future studies are discussed.

Full text

Running Head: Biogeographic Ancestry and Socioeconomic Outcomes
Biogeographic Ancestry and Socioeconomic Outcomes in the Americas: A Meta-Analysis
Emil O. W. Kirkegaard
1
Ulster Institute for Social Research, London, UK
Mingrui Wang
Department of Biology, Beijing University of Agriculture, China
John Fuerst
Ulster Institute for Social Research, London, UK
Narrative reports suggest that socioeconomic status (SES) is associated with biogeographic ancestry (BGA) in
the Americas. If so, SES potentially acts as a confound that needs to be taken into account when evaluating the
relation between medical outcomes and BGA. To explore how systematic BGA-SES associations are, a meta-
analysis of American studies was conducted. 40 studies were identified, yielding a total of 64 independent
samples with directions of associations, including 48 independent samples with effect sizes. An analysis of
association directions found a high degree of consistency. The square root n-weighted directions were 0.83 (K =
36), -0.81 (K = 41) and -0.82 (K = 39) for European, Amerindian and African BGA, respectively. An analysis of
effect size magnitudes found that European BGA was positively associated with SES, with a meta-analytic effect
size of r = .18 [95% CI: .13 to .24, K = 28, n = 35,476.5], while both Amerindian and African BGA were negatively
associated with SES, having meta-analytic effect sizes of -.14 [-.18 to -.10, K = 31, n = 28,937.5] and -.11 [-.15 to
-.07, K = 28, n = 32,710.5], respectively. There was considerable cross-sample variation in effect sizes (mean I2 =
92%), but the sample size was not enough for performing credible moderator analysis. Implications for future
studies are discussed.
Keywords: Biogeographic ancestry, admixture, self-identified race/ethnicity, SES, medical outcomes
Admixture analysis is a potent tool for the exploration of the etiology of traits and trait differences in admixed
populations. Admixture mapping is a form of admixture analysis that allows for the detection of specific disease- and
trait-associated genes (Shriner, 2013; Winkler, Nelson & Smith, 2010). When biogeographic ancestry (BGA) groups
differ in the frequency of disease- or trait-causing genetic variants, in admixed populations the phenotype of interest will
be correlated with BGA in genomic regions near the causal genetic variants, allowing for the identification of associated
loci. Such associations will occur within self-identified race/ethnicity (SIRE) groups. When the genetic architecture of a
trait is complex, with hundreds of loci assumed to contribute to the phenotype, an analysis of global admixture is an
appropriate first step. This strategy seeks to identify associations of global admixture proportions with the measured
trait, without attempting to identify individual loci. Unlike admixture mapping, which requires sample sizes of many
thousands, global admixture analysis can produce meaningful results with smaller sample sizes.
1
Corresponding author, email: emil@emilkirkegaard.dk

2
As used here, population refers to geographically delineated population. In contrast, BGA refers to genetic ancestry
as defined by the presence of ancestrally informative molecular markers in a person’s genome, which allow the
calculation of an individual’s resemblance to the average genotypes in the ancestral populations. These ancestral
populations have also been called clusters or ancestral groups (e.g., Frudakis, 2010; Frudakis & Shriver, 2003; Shriver
& Kittles, 2004). With respect to studies of American populations, the ancestral populations are typically Europeans,
Sub-Saharan Africans, and Amerindians (Salzano & Sans, 2014).
Socioeconomic status (SES) inequalities between and within SIRE groups can lead to spurious associations of BGA
with medical outcomes when both are associated with SES (Marden et al., 2016). Thus, genetic traits associated with
SES whose allele frequencies differ among ancestral groups can be misidentified as being associated with a specific
medical outcome in admixture mapping. For this reason, controls for SES are not infrequently incorporated into such
analyses. Narrative reports (e.g., González Burchard et al., 2005) have suggested that SES covaries with admixture
such that, relative to European BGA, Amerindian and African BGA are associated with lower SES. If this is typically the
case, it would be prudent for researchers to include reliable measures of SES as covariates in analyses to provide
lower-bound estimates of the BGA-medical outcome associations. Moreover, it would be advisable to investigate the
causal pathways mediating the BGA-SES associations, to identify possible unobserved non-genetic mediators of BGA-
medical outcomes associations. However, no meta-analysis has been conducted to date to establish whether SES
outcomes are associated with BGA in any consistent way.
1. Methods
1.1. Collection of studies and data exclusion
Papers were sought which reported quantitative or qualitative associations between European, African, or Amerindian
BGA and socioeconomic outcomes. In 2014/15, a literature search was conducted using the PubMed and BIOSIS
previews databases. Searches such as the following were used: ‘(admixture) AND (socioeconomic or education or
income or SES or poverty) AND (African or European or Amerindian)’. A search, limited to the years 2003 to early 2016,
was also conducted using Google Scholar. For the latter, phrases such as ‘(ses, OR income, OR education) AND
(genomic OR biogeographical, OR ancestry, OR admixture) AND (African, OR European, OR Amerindian)’ were used.
The Google Scholar search yielded over twenty thousand hits in descending order of relevance to the search terms; the
first 1,500 abstracts were skimmed and approximately 250 papers were identified as potential sources and read. The
references of the identified papers, from the PubMed, BIOSIS, and Google Scholar searches, were inspected for
additional sources. In total, we identified 54 papers (51 from the database search and three from the reference search).
Of these, six used redundant samples and thus were excluded. For eight of the papers, while SES was a covariate in an
admixture analysis, no usable information was recoverable and authors did not reply to emails sent to their listed emails.
The remaining 40 papers either contained directions of associations or these were provided by the authors when
contacted; for 27 of the 40 papers, either effect sizes (or metrics that could be converted into effect sizes) were reported
or provided by the authors. Figure 1 shows a flowchart of the procedure.

3
Figure 1. Flowchart of study inclusion.

4
The identified studies and reasons for exclusions are listed in Supplementary File 2. Relevant information from each
study and sample was recorded and is reported in Supplementary File 1. With respect to coding, two of this paper's
authors reviewed studies and reached agreement on ambiguous cases. Many studies did not report effect sizes or
statistics convertible to effect sizes; some studies failed to clearly report association directions. For studies that did not
report effect sizes or directions of associations, authors were contacted. Contact dates, details, and notes are reported
in Supplementary File 1 along with annotated replies when forthcoming. Seventy-three percent of our requests for data
elicited a reply. For 38 percent of requests the authors provided results not reported in the original papers.
Most papers were published in the last few years, with the median year of publication being 2015 (range: 1988 to
2016). Many studies had medical themes and often included both case and control samples. Table 1 shows the
independent samples broken down by sample type. Samples were coded as 'medical' versus 'non-medical' based on
the authors' research goal. As a result, for some samples, this classification was somewhat arbitrary. For example, Zou
et al. (2015) explored genetic assortative mating using the Genes-Environments and Admixture in Latino Americans
(GALA-II) community survey. Since the outcome of interest was 'non-medical', the samples were classified as 'non-
medical'. However, if the same authors had instead explored the relation between BGA and hypertension, using the
same samples, and if results were undifferentiated by clinical conditions, the samples would have been classified as
'case and control' (meaning samples which include both cases and controls).
Table 1. Independent samples by sample type
Sample type
Direction of associations
available
Effect sizes available
Case
10
8
Case and control
24
19
Control
8
4
Non-medical
22
17
Total
64
48
For the meta-analysis, we report results based on associations for which the relevant BGA proportions are above
5%. For comparison, we also note the results with low-admixture samples included. Associations in cases of low
admixture were excluded because when means and thus variances of admixture are low, effect sizes will be attenuated.
Regarding unreliability, one concern is that the present meta-analysis is based on correlation coefficients. As such, the
statistical effects of other BGA components are not held constant and covariance between components can confound
associations, leading to spurious results. Possible confounding is more of a concern when admixture fractions are low.
With this exclusion, the total number of independent samples (64 for directions of associations and 48 for effect sizes)
remained the same, but the number of samples when decomposed by specific BGAs was reduced. The number of
samples is shown in Table 2.
Table 2. Number of independent associations by BGA without (w/o) and with (w) low-admixture samples.

5
Independent
samples
Associations
(African)
Associations
(Amerindian)
Associations
(European)
Directional associations
(w/o low admix)
64
39
41
36
Directional associations (w
low admix)
64
43
43
36
Effect sizes (w/o low admix)
48
28
31
28
Effect sizes (w low admix)
48
31
33
28
Only a subset of the independent samples reported associations for all three BGAs. As a result, the number of
associations for a particular BGA is less than the number of independent samples. When studies reported associations
between outcomes and one or two BGA components, one could attempt to estimate association directions and effect
sizes for the remaining BGAs since the three BGAs are directly related to one another, with ancestries adding up to
unity. For instance, Bonilla (2015, extra information provided by the author) reported that European and Amerindian
BGA correlated at -.82 and -.85 in two samples (ns = 148, 164). The correlations between European BGA and SES
were .10 and .14, respectively; thus, one might infer that the Amerindian × SES correlations were -.10 and -.14. In this
case, however, they were -.01 and -.13; the departure from expectation resulted from the association between African
BGA and SES. To avoid estimation error, unreported associations and effect sizes were not estimated.
1.2. Descriptive statistics of studies
In some instances, associations with multiple outcome measures were reported for a single sample and the same BGA.
Given this situation, treating BGA × outcome associations as independent data points would lead to double counting.
Two recent meta-analyses encountered an analogous problem (Tucker-Drob & Bates, 2015; Tate & McDaniel, 2008).
The method employed by the first was to use a complex weighting approach to avoid double counting, while Tate and
McDaniel (2008) used a simpler approach of averaging results within samples before aggregating. An approach similar
to the second was implemented in the present study. Median values, which are more robust to outliers than are means,
were taken across outcome values for the same BGA within each sample before meta-analyzing the sample
associations. Different classes of outcomes (e.g., education vs. income) were not weighted differently. As Tate and
McDaniel (2008) point out, this method slightly throws off the standard errors, but this problem was judged to be minor.
Henceforth, all results are reported for independent samples (with the exception of Table 3), with BGA × outcome
associations having been first averaged within each sample.
For the 64 independent samples, there were 179 BGA × outcome associations, since some samples included data
for multiple BGAs and multiple SES indicators. Studies reported a large variety of specific outcome variables (Table 3).
As shown in Supplementary File 1, these variables were coded into broad categories. Supplementary Table 1 shows the
breakdown of all outcomes by broad category. Most measures were of individual education, income or occupation. A
few BGA × SES associations were based on the socioeconomic level of the participant's neighborhood. Depending on
the model of the proposed covariance, this could be a questionable index. For example, a recent study based on the UK
Biobank (N ≈ 112k) found a modest phenotypic correlation between individual and neighborhood-level SES (.24) but a

6
strong genetic one (.87) (Hill et al., 2016). While neighborhood-level measures are included in the meta-analysis, it is
advisable that future investigators conduct moderator analyses to estimate the impact of using neighborhood versus
individual indicators. Additionally, several studies reported parental SES. Because most children are the biological
offspring of both of their parents, their admixture will index the average of their parents, and these results can then be
seen as showing the correlation between the parents' SES and the parents' BGA. Roughly sixty percent of the 64
independent samples came from the US. Studies often divided their samples into SIRE groups. Tables 4a and 4b show
a breakdown by country and SIRE group.
Table 3. Broad outcomes (specific outcomes) for all BGA × outcome associations with/without low admixture excluded
Broad outcome (specific outcomes)
Frequency
Broad outcome (specific outcomes)
Frequency
Education (education, years of schooling)
69/73
Parental SES (parental Hollingshead
index, running water, had a car)
6/6
Income (income, wealth index, household
income per capita)
49/53
Parental education
2/2
SES (SES, SES + education, occupation,
general factor of SES, Hollingshead index,
Yost index, home property band, household
asset index)
34/35
Parental income
1/1
Neighborhood SES
18/19
Total
179/189
Table 4a. Number of independent samples and sample sizes for directional association analysis by SIRE group and
country. Low admixture excluded.
SIRE group
K
N
SIRE group
K
N
SIRE group
K
N
African American (US)
10
20581
Multi-ethnic (Chile)
2
1967
African-descent (T&T)
1
107
Hispanic (US)
10
9983
Multi-ethnic (Mexico)
2
1991
Multi-ethnic (Argentina)
1
220
Mexican (US)
5
1050
Multi-ethnic (US)
2
2031
Multi-ethnic (Colombia)
1
2092
Multi-ethnic (Brazil)
6
6696
Mestizo (Chile)
2
2381
Not stated (Costa Rica)
1
1998
Puerto Rican (PR)
7
2368
Native American (US)
2
741
Puerto Rican (US)
1
122
Not stated (Mexico)
3
1831
Not stated (Uruguay)
2
312
White American (US)
1
73
Multi-ethnic (Peru)
3
1979
Not stated (Colombia)
2
1971
Total
64
Note: N = participant sample size; K = number of independent samples. PR = Puerto Rico; T&T = Trinidad and
Tobago.

7
Table 4b. Number of independent samples and sample sizes for effect size analysis by SIRE group and country. Low
admixture excluded.
SIRE group
K
N
SIRE group
K
N
SIRE group
K
N
African American (US)
7
13589
Multi-ethnic (Chile)
2
1967
African-descent (T&T)
0
Hispanic (US)
6
8041
Multi-ethnic (Mexico)
2
1991
Multi-ethnic (Argentina)
1
220
Mexican (US)
5
1050
Multi-ethnic (US)
2
2031
Multi-ethnic (Colombia)
1
2092
Multi-ethnic (Brazil)
5
6518.5
Mestizo (Chile)
0
Not stated (Costa
Rica)
0
Puerto Rican (PR)
3
1943
Native American
(US)
2
741
Puerto Rican (US)
1
122
Not stated (Mexico)
3
1831
Not stated (Uruguay)
2
312
White American (US)
1
73
Multi-ethnic (Peru)
3
1979
Not stated
(Colombia)
2
1971
Total
48
1.3. Directions of associations
As shown in Table 2, directions of associations were more frequently available than were actual effect sizes.
Association directions can provide some indication as to whether the findings are in line with a null hypothesis. Other
meta-analyses have included analyses of association directions (e.g., Van der Meer & Tolsma, 2014).
1.4. Statistical analysis
For each sample, directions (negative, null, positive) were assigned to associations. With respect to this assignment,
neither p-values nor the magnitude of effects were taken into account. Thus, for example, the correlation between
African BGA and education for Bonilla et al.’s (2015) sample 1 was -.16 (n = 164, p = .08), while the correlation between
African BGA and education for Bonilla et al.’s (2015) sample 2 was .04 (n = 148, p = .59); the first association was
coded as 'negative' and the second as 'positive' despite the difference in effect size. Those samples coded as 'null' were
ones where the effect size was zero. After directions were assigned, they were coded as negative = –1, null = 0, and
positive = 1). If a sample had consistent directions (across multiple SES indicators), it received a score of either -1 or 1.
If a sample had mixed results, it received an intermediate score equivalent to the mean of the different associations
(e.g., a sample reporting three positive and one negative associations received a score of (3-1)/4=0.5.) The results are
reported unweighted; weighted by the square root of the sample sizes; and weighted by the sample sizes. We consider
square root n-weighted results to be preferable as they strike a good balance between taking into account the effect of
sample size and not obscuring results from smaller samples; other researchers have employed this weighting strategy
(for example, Kan et al., 2013).
2. Results
2.1. Directions of associations

8
The aggregated within-sample directions are shown in Table 5. In general, the directions are positive for European BGA
and negative for the other two ancestries. With respect to African and Amerindian BGA, the weighted results are
substantially stronger when low-admixture samples are excluded. There were no samples with low European BGA.
Table 5. Directions of associations by BGA
BGA
N
K
Mean
direction
(unweighted)
Mean direction
(SQRT(n)-
weighted)
Mean direction
(n-weighted)
African (w/o low admix)
40931.5
39
-0.67
-0.82
-0.94
African (w low admix)
47742.5
43
-0.75
-0.64
-0.56
Amerindian (w/o low admix)
35577.5
41
-0.76
-0.81
-0.85
Amerindian (w low admix)
47107.5
43
-0.72
-0.65
-0.41
European (w/o low admix)
39548.5
36
0.72
0.83
0.89
European (w low admix)
39548.5
36
0.72
0.83
0.89
2.2. Effect sizes
For 92% of the samples with effect sizes, Pearson correlations or other effect sizes that could be directly converted into
these (i.e., r2, Spearman's rho, beta coefficients, and odds ratio) were available. For other samples, effect sizes needed
to be computed from t-tests (N = 1), F-tests (N = 4), and frequency tables (N = 5). Conversions were made using the
formulas and methods noted in Supplementary File 1. To provide an overview of the effect sizes, a density histogram
plot was made of the correlations for each BGA component. Additional statistics for the effect sizes were also computed.
For the meta-analysis, because samples were from different populations in different countries, a random effects model
was appropriate (Schmidt & Hunter, 2014). We conducted the analysis using the metafor package for R (Viechtbauer,
2015).
The studies varied substantially in sample size. Figure 2 shows the distribution of the sample sizes for the effect
size analysis with low-admixture samples excluded. The density histogram plot of the effect sizes is shown in Figure 3.
Descriptive statistics for the effect sizes with and without low admixture samples excluded are shown in Table 6.

9
Figure 2. Density histogram of sample sizes, with a stippled line at the mean.

10
Figure 3. Density histogram plot for effect sizes by BGA.
Table 6. Descriptive statistics for the effect sizes
BGA
N
K
Mean
Median
Max
Min
SD
10th
centile
90th
centile
African (w/o low
admix)
32710.5
28
-.11
-.12
.14
-.35
.11
-.21
.03
African (w low
admix)
37523.5
31
-.09
-.10
.14
-.35
.11
-.20
.08
Amerindian (w/o
low admix)
28937.5
31
-.13
-.13
.17
-.39
.12
-.31
.00
Amerindian (w
low admix)
40467.5
33
-.13
-.12
.17
-.39
.12
-.30
.00
European (w/o
low admix)
35476.5
28
.17
.19
.53
-.24
.16
.02
.33

11
European (w low
admix)
35476.5
28
.17
.19
.53
-.24
.16
.02
.33
Note: n = participant sample size; K = number of independent samples. With/without BGA <5%.
There are two clear outliers for European BGA at -.24 and -.20. The first one was based on an extremely small
sample (n = 15). The second one was both an outlier for European BGA (Zou et al. sample 6, N = 158) and for
Amerindian BGA.
2
When these outliers are excluded, the mean correlations changed little (rs = .16, -.14, and -.11 for
European, Amerindian and African BGA respectively). Figures 4, 5 and 6 show the forest plots for the random effects
models. Tables 7a to 7c show study details and effect sizes.
Figure 4. Forest plot for European BGA results, based on random effects model
2
The lead author (James Zou) was contacted to verify the data. He checked the code and data and reported that there were no
errors.

12
Figure 5. Forest plot for Amerindian BGA results, based on random effects model

13
Figure 6. Forest plot for African BGA results, based on random effects model
Table 7a. Study details and effect sizes (with lower and upper 95% confidence intervals) for the meta-analyses:
European BGA.
Author sample
Country
Ethnic group
N
r
Lower
Upper
Adhikari et al. (2016)_1
Brazil
Multi-ethnic
1375
0.02
-0.03
0.07
Adhikari et al. (2016)_2
Chile
Multi-ethnic
1934
0.01
-0.03
0.05
Adhikari et al. (2016)_3
Colombia
Multi-ethnic
2092
0.14
0.09
0.18
Adhikari et al. (2016)_4
Mexico
Multi-ethnic
1535
0.16
0.12
0.21
Adhikari et al. (2016)_5
Peru
Multi-ethnic
1344
0.15
0.10
0.20
Bonilla et al. (2015)_1
Uruguay
Not stated
164
0.22
0.07
0.36
Bonilla et al. (2015)_2
Uruguay
Not stated
148
0.03
-0.13
0.19
Cardel et al. (2012)_1
US
Multi-ethnic
305
0.50
0.40
0.60
Fejerman et al. (2010)_1
Mexico
Not stated
1035
0.16
0.10
0.22
Florez et al. (2009)_1
Mexico
Not stated
235
0.33
0.21
0.45

14
Florez et al. (2009)_2
Colombia
Not stated
696
0.34
0.27
0.41
Klimentidis et al. (2009)_1
US
Hispanic
147
0.12
-0.05
0.28
Klimentidis et al. (2009)_2
US
Native American
15
-0.24
-0.72
0.25
Lima-Costa et al. (2016)_1
Brazil
Not stated
1305.5
0.20
0.15
0.26
Martinez-Marignac et al.
(2007)_1
Mexico
Not stated
561
0.53
0.46
0.60
Menezes et al. (2015)_1
Brazil
Multi-ethnic
2960
0.22
0.18
0.26
Qi et al. (2012)_2
US
Hispanic
4980
0.19
0.16
0.22
Robbins et al. (2015)_1
US
African-American
11239
0.19
0.17
0.20
Schlesinger et al. (2011)_1
Brazil
Not stated
291
0.12
0.01
0.23
Via et al. (2011)_1
US
Puerto-Rican
642
0.16
0.08
0.24
Zou et al. (2015)_1
US
Puerto-Rican
686
0.10
0.03
0.17
Zou et al. (2015)_2
US
Puerto-Rican
615
0.23
0.15
0.30
Zou et al. (2015)_3
US
Mexican
157
0.20
0.05
0.35
Zou et al. (2015)_4
US
Mexican
163
0.20
0.05
0.35
Zou et al. (2015)_5
US
Mexican
291
0.20
0.09
0.31
Zou et al. (2015)_6
US
Mexican
158
-0.20
-0.35
-0.05
Zou et al. (2015)_7
US
Mexican
281
0.22
0.11
0.33
Zou et al. (2015)_8
US
Puerto-Rican
122
0.26
0.09
0.43
RE Model
35476.5
0.18
0.13
0.24
Table 7b. Study details and effect sizes (with lower and upper 95% confidence intervals) for the meta-analyses:
Amerindian BGA.
Author sample
Country
Ethnic group
N
r
Lower
Upper
Adhikari et al. (2016)_1
Brazil
Multi-ethnic
1375
-0.01
-0.06
0.04
Adhikari et al. (2016)_2
Chile
Multi-ethnic
1934
-0.02
-0.07
0.02
Adhikari et al. (2016)_3
Colombia
Multi-ethnic
2092
-0.08
-0.12
-0.03
Adhikari et al. (2016)_4
Mexico
Multi-ethnic
1535
-0.16
-0.21
-0.11
Adhikari et al. (2016)_5
Peru
Multi-ethnic
1344
-0.15
-0.20
-0.10
Aldrich et al. (2013)_3
US
Hispanic
299
-0.12
-0.23
-0.01
Aldrich et al. (2013)_4
US
Hispanic
113
-0.31
-0.49
-0.13
Bonilla et al. (2015)_1
Uruguay
Not stated
164
-0.20
-0.35
-0.04

15
Bonilla et al. (2015)_2
Uruguay
Not stated
148
0.00
-0.16
0.17
Campbell et al. (2012)_1
Colombia
Not stated
1275
-0.19
-0.24
-0.14
Goonesekera et al. (2015)_1
US
Multi-ethnic
1726
-0.08
-0.13
-0.04
Lima-Costa et al. (2016)_1
Brazil
Not stated
1305.5
-0.24
-0.30
-0.19
Menezes et al. (2015)_1
Brazil
Multi-ethnic
2960
-0.17
-0.21
-0.13
Norden-Krichmar et al. (2014)_1
US
Native American
726
-0.07
-0.14
0.00
Pereira et al. (2012)_1
Peru
Multi-ethnic
541
-0.10
-0.18
-0.02
Qi et al. (2012)_2
US
Hispanic
4980
-0.15
-0.18
-0.12
Salam et al. (2015)_1
US
Hispanic
1818
-0.39
-0.43
-0.35
Sánchez et al. (2012)_1
Mexico
Not stated
456
-0.34
-0.43
-0.26
Sánchez et al. (2012)_2
Argentina
Not stated
220
-0.33
-0.46
-0.20
Sánchez et al. (2012)_3
Peru
Not stated
94
-0.10
-0.30
0.10
Sánchez et al. (2012)_4
Chile
Not stated
33
-0.13
-0.48
0.22
Sweeney et al. (2007)_1
US
Hispanic
684
-0.11
-0.18
-0.04
Via et al. (2011)_1
US
Puerto-Rican
642
0.00
-0.08
0.08
Zou et al. (2015)_1
US
Puerto-Rican
686
-0.05
-0.12
0.02
Zou et al. (2015)_2
US
Puerto-Rican
615
-0.05
-0.13
0.03
Zou et al. (2015)_3
US
Mexican
157
-0.20
-0.35
-0.05
Zou et al. (2015)_4
US
Mexican
163
-0.17
-0.32
-0.02
Zou et al. (2015)_5
US
Mexican
291
-0.22
-0.33
-0.11
Zou et al. (2015)_6
US
Mexican
158
0.17
0.02
0.32
Zou et al. (2015)_7
US
Mexican
281
-0.20
-0.31
-0.09
Zou et al. (2015)_8
US
Puerto-Rican
122
0.00
-0.18
0.18
RE Model
28937.5
-0.14
-0.18
-0.10
Table 7c. Study details and effect sizes (with lower and upper 95% confidence intervals) for the meta-analyses: African
BGA.
Author sample
Country
Ethnic group
n
r
Lower
Upper
Adhikari et al. (2016)_1
Brazil
Multi-ethnic
1375
-0.06
-0.11
0.00
Adhikari et al. (2016)_3
Colombia
Multi-ethnic
2092
-0.12
-0.16
-0.07
Aldrich et al. (2013)_1
US
African-American
280
-0.18
-0.29
-0.06
Aldrich et al. (2013)_2
US
African-American
255
-0.23
-0.35
-0.11

16
Blanton et al. (2008)_1
Brazil
Multi-ethnic
587
-0.35
-0.43
-0.27
Bonilla et al. (2015)_1
Uruguay
Not stated
164
-0.12
-0.27
0.04
Bonilla et al. (2015)_2
Uruguay
Not stated
148
-0.05
-0.22
0.11
Bress et al. (2015)_2
US
African-American
95
-0.04
-0.24
0.17
Cardel et al. (2012)_1
US
Multi-ethnic
305
-0.15
-0.26
-0.04
Ducci et al. (2009)_1
US
African-American
228
-0.14
-0.27
-0.01
Goonesekera et al. (2015)_1
US
Multi-ethnic
1726
-0.06
-0.11
-0.02
Gower et al. (2003)_1
US
African-American
52
-0.13
-0.40
0.14
Gower et al. (2003)_2
US
White-American
73
-0.08
-0.31
0.15
Lima-Costa et al. (2016)_1
Brazil
Not stated
1305.5
-0.16
-0.21
-0.11
Menezes et al. (2015)_1
Brazil
Multi-ethnic
2960
-0.19
-0.23
-0.15
Qi et al. (2012)_2
US
Hispanic
4980
-0.07
-0.10
-0.04
Robbins et al. (2015)_1
US
African-American
11239
-0.19
-0.20
-0.17
Schlesinger et al. (2011)_1
Brazil
Not stated
291
-0.15
-0.26
-0.03
Tandon et al. (2015)_1
US
African-American
1440
-0.10
-0.15
-0.04
Via et al. (2011)_1
US
Puerto-Rican
642
-0.17
-0.25
-0.09
Zou et al. (2015)_1
US
Puerto-Rican
686
-0.05
-0.12
0.02
Zou et al. (2015)_2
US
Puerto-Rican
615
-0.20
-0.28
-0.13
Zou et al. (2015)_3
US
Mexican
157
0.10
-0.06
0.26
Zou et al. (2015)_4
US
Mexican
163
-0.10
-0.25
0.05
Zou et al. (2015)_5
US
Mexican
291
0.14
0.03
0.25
Zou et al. (2015)_6
US
Mexican
158
0.10
-0.06
0.26
Zou et al. (2015)_7
US
Mexican
281
0.00
-0.12
0.12
Zou et al. (2015)_8
US
Puerto-Rican
122
-0.28
-0.45
-0.11
RE Model
32710.5
-0.11
-0.15
-0.07
The random effect meta-analytic mean effect sizes are .18, -.14, and -.11 for European, Amerindian and African
BGA, respectively. European BGA shows a stronger association with SES than do Amerindian and African BGA,
presumably because it is the largest ancestry component (mean European BGA = 54.41%). The difference could be due
to the higher variation in the proportion of European BGA in the subjects. The mean standard deviation of admixture
(18.01, 11.18, and 14.50 for European, Amerindian and African, respectively) was lower for the non-European
ancestries, and less variation is expected to lead to smaller effect sizes owing to restriction of range (Schmidt & Hunter,
2014).
The exclusion of samples with a mean BGA below 5% had only a slight impact on the results. When all samples are
included, the meta-analytic effect sizes are r = .18 [.13, .24], r = -.13 [-.17, -.09], and r = -.09 [-.13, -.05] for European,

17
Amerindian and African BGA, respectively. Results for African BGA are impacted the most owing to the presence of a
positive association between African BGA and outcomes for several samples with very low African admixture (e.g.,
Adhikari et al., 2016: Chile, African admixture 2.61%, r BGA × SES = .08). It is notable that for two of the same
countries, namely Chile and Mexico, Fuerst and Kirkegaard (2016a, 2016b) found that regional African BGA is
associated with higher SES.
The funnel plots did not show notable asymmetry, suggesting no publication bias. Figures 7, 8, and 9 show,
respectively, the funnel plots for European, Amerindian, and African BGA.
Figure 7. Funnel plot for European BGA, with standard error on the y-axis and effect size on the x-axis

18
Figure 8. Funnel plot for Amerindian BGA, with standard error on the y-axis and effect size on the x-axis

19
Figure 9. Funnel plot for African BGA, with standard error on the y-axis and effect size on the x-axis
3. Discussion
3.1. Main results
The results of the meta-analysis are consistent with those reported in earlier narrative reviews in that European BGA
was statistically associated with more favorable socioeconomic outcomes relative to Amerindian and African BGA.
When low-admixture samples were excluded, the random effects meta-analysis gave a mean SES × BGA correlation of
r = .18, r = -.14, and r = -.11 for European, Amerindian and African BGA, respectively; furthermore, the analysis of SES
× BGA association directions gave square root n-weighted means of M = 0.83, M = -0.81, and M = -0.82 for European,
Amerindian and African BGA, respectively. For the random effects meta-analysis, the heterogeneity, or the percentage
of variance due to variance between samples as opposed to sampling error, was considerable (I2 mean = 92%, I2 European =
95%, I2 Amerindian = 91%, I2 African = 89%) by conventional standards (Higgins & Green, 2008). Due to the limited number of
samples, it is difficult to evaluate the cause of this pattern of results. Some possibilities are as follows:
1. Discretization: Many outcome variables were ordinal instead of continuous, even when continuous values were
available (e.g., income). Pearson correlations assume that the data are normally and continuously distributed, so the
use of non-continuous data induces a downward bias in the results. Corrections for discretization were not attempted,
although one could attempt to apply these (Schmidt & Hunter, 2014).

20
2. Number of BGA-informative markers: There were large differences in the number of genomic markers used to
estimate individual BGA. Figure 10 shows a density histogram of the distribution for the effect size analysis with low-
admixture samples removed. Using fewer markers results in more measurement error with respect to true BGA (Russo
et al., 2016). Ruiz-Linares et al. (2014) reported correlations between admixture estimates using different numbers of
markers. They noted that a recent study (Scharf et al., 2013) had found that using 15 markers resulted in correlations of
about .60 with estimates derived from 50k markers. Furthermore, using 30 and 152 markers resulted in r = .70 and r =
.85, respectively, with respect to estimated admixture based on 50k markers. It is clear that there are diminishing returns
to using more markers, but also that using more reduces measurement error with respect to true BGA. It would be
difficult to correct for measurement error with respect to BGA. One option would be to acquire a sufficient number of
data points from one study to allow for the estimation of a predictive model. One could then use that model to estimate
the measurement error in other studies based on the number of reported markers used. However, we failed to find a
study which had sufficient data points, so a correction was not attempted.
Figure 10. Density histogram of the number of genetic markers in each sample. Average frequency density on the y-
axis and number of genetic markers on the x-axis.

21
3. Variation in admixture: Based on the samples for which standard deviations of BGA were reported, variability in BGA
differed considerably between samples. For instance, for Uruguayans, Bonilla et al. (2015) found a standard deviation of
African BGA of 7.52; in contrast, for Brazilians, Schlesinger et al. (2011) found one of 24.08. As mentioned earlier, the
magnitude of associations between outcomes and BGA will be impacted by variance in BGA. Thus, between-sample
heterogeneity in BGA variance is expected to contribute to heterogeneity in BGA × SES associations.
4. Heterogeneous origin: The studies in the meta-analysis came from many countries. It is probable that the statistical
effect of one's BGA depends on local cultural norms or practices that differ between countries or even between regions
within countries.
5. Sampling: Most of the studies used convenience samples, an approach that can have an impact on associations.
Adhikari et al.’s (2016) samples, for example, were recruited mostly at universities, meaning that individuals were
selected with respect to educational status. Thus, the correlations they obtained are expected to be attenuated relative
to those derived from more representative samples.
3.2. Implications for epidemiological studies
Fairly robust associations between BGA and SES were found. Given this, to avoid spurious associations in regression
analyses due to omitted variable bias, it is potentially important to include reliable measures of SES in studies of
medical outcomes. Studies reviewed in this analysis employed a wide variety of specific measures (see Table 3), some
of which may be better indices of medically relevant SES than others. It would be worthwhile to investigate which
measures (neighborhood SES, education, income, etc.) are generally associated to a greater degree with BGA and with
medical outcomes, so that future studies could attempt to incorporate these measures into analyses.
Further, it is also important to identify the factors mediating the BGA-SES associations, as these could incrementally
explain the BGA-medical outcome associations. Figure 11 shows one possible model. In this model, local ancestry is a
subset of global ancestry. Both local and global ancestry correlate with disease loci which, in turn, are causally related
to health outcomes. Global ancestry is also related both to an individual's SIRE and to cultural practices associated with
SIRE, and to an individual's race/ethnicity-associated phenotype, and social reactions, perhaps in the form of “colorism”
(e.g., Hunter, 2007; Telles, 2014), associated with that phenotype. In this model, both direct and indirect pathways run
from cultural practices and social reactions to health outcomes. An example of a direct pathway would be through
observed race/ethnicity-related differential treatments by providers (as discussed by, e.g., Van Ryn, 2002). Indirectly,
these factors potentially impact both individuals' human capital and SES, which are reciprocally related to one another
and to health outcomes.

22
Figure 11. Model of possible relations between BGA and SES. The stippled lines represent correlations and the
arrowed lines represent causal associations.
This model is similar to that proposed by Marden et al. (2016). A notable exception is that we include human capital
as a mediating variable. Cognitive epidemiology studies have shown that measures of cognitive ability (one component
of human capital) can explain a portion of medical-related outcomes, independent of SES (Deary, 2009; Der, Batty &
Deary, 2009; Wraw et al., 2015). Because, throughout the Americas, SIRE groups are known to differ in mean levels of
phenotypic cognitive ability (Fuerst & Kirkegaard, 2016b; Roth et al., 2001), it would be reasonable to include measures
of cognitive ability in admixture analyses to see if such measures explain some of the health outcome differences
independent of SES. However, we could not find any study that did this. On the other hand, we did locate one recent
admixture paper (Akshoomoff et al., 2014) which reported that BGA was associated with cognitive ability. On inquiry, as
expected given the SIRE differences, European BGA was positively associated with cognitive ability, while African and
Amerindian BGA were negatively associated (Akshoomoff, personal communications, November 9, 2014). If cognitive
ability mediates the relationship between BGA and SES, not including measures of it constitutes omitted variable bias.
3.3. Untangling the statistical effects of BGA and SIRE
In many of the samples included in this meta-analysis, individual SES outcomes are associated with BGA within SIRE
groups. Hence, SIRE membership does not mediate the relationship. In other cases, particularly in Latin America, the
issue is less clear and BGA may be confounded with SIRE. For instance, Leite et al. (2011) found that European BGA
was positively correlated with SES in Brasília. This could be because SES was positively correlated with BGA
independently of SIRE, or because it was positively correlated with SIRE but not with BGA independent of SIRE. The
analysis by Ruiz-Linares et al. (2014) has helped to clarify the issue. The authors looked at the association between
BGA, skin pigmentation and SIRE in a multi-country sample from Brazil, Chile, Colombia, Mexico and Peru (mean age
20 to 25, depending on country). The authors found modest correlations between BGA and SIRE (e.g., 0.48 in the case

23
of both European and Amerindian BGA/SIRE). They reported that wealth and educational attainment correlated with
European BGA (r = .12 for the full sample). However, when the statistical effect of BGA was controlled for, education
was not significantly associated with SIRE, and wealth was only related to SIRE within the European/White color group.
Similar results in this regard were found with Menezes et al.'s (2015) sample from the 1982 Pelotas Birth Cohort study
(F. Hartwig, personal communication, March 4, 2016). Details pertaining to this sample can be found in Menezes et al.
(2015). The results pertaining to self-identified color and interviewer-rated color are shown in Supplementary File 3. For
this sample, the positive association between European BGA and household assets, schooling, and income was robust
to controls for interviewer-observed race/ethnicity and SIRE. Contrary to a simple colorism model, Black (preta) and
Brown (parda) interviewer-rated race/ethnicity were generally positively related to outcomes, controlling for European
BGA. Taken together, the results suggest that the association between BGA and SES is substantially independent of
SIRE and thus SIRE-related cultural factors. Quite possibly, phenotype-based social reactions, which would differentially
impact individuals within SIRE groups, affect outcomes. For example, consistent with this possibility, Telles (2014)
found, in a large multinational sample, that interviewer-rated skin brightness was related to SES controlling for SIRE.
3.4. Limitations and suggestions for future research
The number of effect sizes used in the present meta-analysis is limited. This is because many studies did not report
effect sizes and, in some cases, the authors either did not reply to emails or were unable to provide results or data.
Meta-analyses would be more reliable and easier to perform if scientists were more willing to publish their data and
report effect sizes. The results significantly varied across studies (mean effect size heterogeneity = 92%), which means
that there are effect size moderators. A moderator analysis was not conducted due to the relatively small number of
studies in the dataset. We suggest that moderator analyses be conducted as relevant data accumulates. Detailed
analyses for publication bias were also not conducted because the number of effect sizes was too small for a reliable
analysis. This concern should be investigated in future meta-analyses.
Supplementary material
Supplementary File 1-3, high-quality figures and R analysis code are available at the repository at Open Science
Framework https://osf.io/ydc3f/files/.
Conflicts of Interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that
could be construed as a potential conflict of interest.
References
* indicates that the study included data used in the meta-analysis.
*Adhikari, K., Fontanil, T., Cal, S., Mendoza-Revilla, J., Fuentes-Guajardo, M., Chacón-Duque, J.C., ... & Jaramillo, C.
(2016). A genome-wide association scan in admixed Latin Americans identifies loci influencing facial and scalp hair
features. Nature Communications, 7: 10815.
Akshoomoff, N., Newman, E., Thompson, W.K., McCabe, C., Bloss, C.S., Chang, L., Amaral, D.G. et al. (2014). The
NIH Toolbox Cognition Battery: Results from a large normative developmental sample (PING). Neuropsychology 28(1):
1.

24
*Aldrich, M.C., Selvin, S., Wrensch, M.R., Sison, J.D., Hansen, H.M., Quesenberry Jr, C.P. ... & Wiencke, J.K. (2013).
Socioeconomic status and lung cancer: Unraveling the contribution of genetic admixture. American Journal of Public
Health 103(10): e73-e80.
*Allison, M.A., Peralta, C.A., Wassel, C.L., Aboyans, V., Arnett, D.K., Cushman, M. ... & Criqui, M.H. (2010). Genetic
ancestry and lower extremity peripheral artery disease in the Multi-Ethnic Study of Atherosclerosis. Vascular Medicine
15(5): 351-359.
*Blanton, R.E., Silva, L.K., Morato, V.G., Parrado, A.R., Dias, J P., Melo, P.R. ... & Teixeira, M.G. (2008). Genetic
ancestry and income are associated with dengue hemorrhagic fever in a highly admixed population. European Journal
of Human Genetics 16(6): 762-765.
*Bonilla, C., Bertoni, B., Hidalgo, P.C., Artagaveytia, N., Ackermann, E., Barreto, I., Cancela, P. et al. (2015). Breast
cancer risk and genetic ancestry: A case–control study in Uruguay. BMC Women's Health 15(1): 1.
*Bress, A., Kittles, R., Wing, C., Hooker Jr, S.E. & King, A. (2015). Genetic ancestry as an effect modifier of naltrexone
in smoking cessation among African Americans: An analysis of a randomized controlled trial. Pharmacogenetics and
Genomics 25(6): 305-312.
*Campbell, D.D., Parra, M.V., Duque, C., Gallego, N., Franco, L., Tandon, A. ... & McCarthy, M.I. (2012). Amerind
ancestry, socioeconomic status and the genetics of type 2 diabetes in a Colombian population. PLoS One 7(4): e33570.
*Cardel, M., Willig, A.L., Dulin-Keita, A., Casazza, K., Beasley, T.M. & Fernández, J.R. (2012). Parental feeding
practices and socioeconomic status are associated with child adiposity in a multi-ethnic sample of children. Appetite
58(1): 347-353.
*Cheng, C.Y., Reich, D., Haiman, C.A., Tandon, A., Patterson, N., Elizabeth, S. ... & Altshuler, D. (2012). African
ancestry and its correlation to type 2 diabetes in African Americans: A genetic admixture analysis in three US population
cohorts. PLoS One 7(3): e32840.
*Choudhry, S., Burchard, E.G., Borrell, L.N., Tang, H., Gomez, I., Naqvi, M. ... & Ziv, E. (2006). Ancestry–environment
interactions and asthma risk among Puerto Ricans. American Journal of Respiratory and Critical Care Medicine
174(10): 1088-1093.
*Cifuentes, L., Morales, R., Sepulveda, D., Jorquera, H. & Acuna, M. (2004). DYS19 and DYS199 loci in a Chilean
population of mixed ancestry. American Journal of Physical Anthropology 125(1): 85-89.
Deary, I.J. (2009). Introduction to the special issue on cognitive epidemiology. Intelligence 37: 517-519.
Der, G., Batty, G.D. & Deary, I.J. (2009). The association between IQ in adolescence and a range of health outcomes at
40 in the 1979 US National Longitudinal Study of Youth. Intelligence 37: 573-580.
*Ducci, F., Roy, A., Shen, P.H., Yuan, Q., Yuan, N.P., Hodgkinson, C.A. ... & Goldman, D. (2009). Association of
substance use disorders with childhood trauma but not African genetic heritage in an African American cohort. American
Journal of Psychiatry 166: 1031-1040.
*Fejerman, L., Hu, D., Huntsman, S., John, E.M., Stern, M.C., Haiman, C.A. ... & Ziv, E. (2013). Genetic ancestry and
risk of mortality among US Latinas with breast cancer. Cancer Research 73: 7243-7253.

25
*Fejerman, L., Romieu, I., John, E.M., Lazcano-Ponce, E., Huntsman, S., Beckman, K.B. ... & Torres-Mejía, G. (2010).
European ancestry is positively associated with breast cancer risk in Mexican women. Cancer Epidemiology Biomarkers
& Prevention 19: 1074-1082.
*Florez, J.C., Price, A.L., Campbell, D., Riba, L., Parra, M.V., Yu, F. ... & Franco, L. (2009). Strong association of
socioeconomic status with genetic ancestry in Latinos: Implications for admixture studies of type 2 diabetes.
Diabetologia 52: 1528-1536.
Frudakis, T. (2010). Molecular Photofitting: Predicting Ancestry and Phenotype Using DNA. Elsevier.
Frudakis, T. & Shriver, M. (2003). Compositions and methods for inferring ancestry. U.S. Patent Application 10/644,594,
filed August 19, 2003.
Fuerst, J. & Kirkegaard, E.O.W. (2016a). Admixture in the Americas: Regional and national differences. Mankind
Quarterly 56: 255-373.
Fuerst, J. & Kirkegaard, E.O.W. (2016b). The genealogy of differences in the Americas. Mankind Quarterly 56: 425-481.
*González Burchard, E., Borrell, L.N., Choudhry, S., Naqvi, M., Tsai, H.-J., Rodriguez-Santana, J.R., Chapela, R. et al.
(2005). Latino populations: A unique opportunity for the study of race, genetics, and social environment in
epidemiological research. American Journal of Public Health 95: 2161-2168.
*Goonesekera, S.D., Fang, S.C., Piccolo, R.S., Florez, J.C. & McKinlay, J.B. (2015). Biogeographic ancestry is
associated with higher total body adiposity among African-American females: The Boston area community health
survey. PloS One 10(4): e0122808.
*Gower, B.A., Fernández, J.R., Beasley, T.M., Shriver, M.D. & Goran, M.I. (2003). Using genetic admixture to explain
racial differences in insulin-related phenotypes. Diabetes 52: 1047-1051.
*Gravlee, C.C., Non, A.L. & Mulligan, C.J. (2009). Genetic ancestry, social classification, and racial inequalities in blood
pressure in Southeastern Puerto Rico. PLoS One 4(9): e6821.
Higgins, J.P.T & Green, S., eds. (2008). Cochrane Handbook for Systematic Reviews of Interventions, Vol. 5.
Chichester: Wiley-Blackwell.
Hill, W.D., Hagenaars, S.P., Marioni, R.E., Harris, S.E., Liewald, D.C.M., Davies, G., McIntosh, A.M., Gale, C.R., Deary,
I.J. & International Consortium for Blood Pressure (2016). Molecular genetic contributions to social deprivation and
household income in UK Biobank (n= 112,151). bioRxiv: 043000.
Hunter, M. (2007). The persistent problem of colorism: Skin tone, status, and inequality. Sociology Compass 1(1): 237-
254.
Kan, K.-J., Wicherts, J.M., Dolan, C.V. & van der Maas, H.L.J. (2013). On the nature and nurture of intelligence and
specific cognitive abilities. The more heritable, the more culture dependent. Psychological Science 24: 2420-2428.
*Klimentidis, Y.C., Miller, G.F. & Shriver, M.D. (2009). The relationship between European genetic admixture and body
composition among Hispanics and Native Americans. American Journal of Human Biology 21: 377-382.

26
*Leite, T.K.M., Fonseca, R.M.C., De França, N.M., Parra, E.J. & Pereira, R.W. (2011). Genomic ancestry, self-reported
“color” and quantitative measures of skin pigmentation in Brazilian admixed siblings. PLoS One 6(11): e27162.
*Lima-Costa, M.F., de Mello Mambrini, J.V., Leite, M.L.C., Peixoto, S.V., Firmo, J.O.A., de Loyola Filho, A.I. ... &
Tarazona-Santos, E. (2016). Socioeconomic position, but not African genomic ancestry, is associated with blood
pressure in the Bambui-Epigen (Brazil) Cohort Study of Aging. Hypertension 67: 349-355.
Marden, J.R., Walter, S, Kaufman, J.S. & Glymour, M.M. (2016). African ancestry, social factors, and hypertension
among non-Hispanic Blacks in the Health and Retirement Study. Biodemography and Social Biology 62: 19-35.
*Martinez-Marignac, V.L., Valladares, A., Cameron, E., Chan, A., Perera, A., Globus-Goldberg, R. ... & Shriver, M.D.
(2007). Admixture in Mexico City: Implications for admixture mapping of type 2 diabetes genetic risk factors. Human
Genetics 120: 807-819.
*Menezes, A.M.B., Wehrmeister F.C., Hartwig, F.P., Perez-Padilla, R., Gigante, D.P., Barros, F.C., Oliveira, I.O.,
Ferreira, G.D. & Horta, B.L. (2015). African ancestry, lung function and the effect of genetics. European Respiratory
Journal: ERJ-01121.
*Molokhia, M., Hoggart, C., Patrick, A.L., Shriver, M., Parra, E., Ye, J. & McKeigue, P.M. (2003). Relation of risk of
systemic lupus erythematosus to West African admixture in a Caribbean population. Human Genetics 112: 310-318.
*Norden‐Krichmar, T.M., Gizer, I.R., Libiger, O., Wilhelmsen, K.C., Ehlers, C.L. & Schork, N.J. (2014). Correlation
analysis of genetic admixture and social identification with body mass index in a Native American community. American
Journal of Human Biology 26: 347-360.
*Pereira, L., Zamudio, R., Soares-Souza, G., Herrera, P., Cabrera, L., Hooper, C.C. ... & Schneider, S. (2012).
Socioeconomic and nutritional factors account for the association of gastric cancer with Amerindian ancestry in a Latin
American admixed population. PLoS One 7(8): e41200.
*Qi, L., Nassir, R., Kosoy, R., Garcia, L., Curb, J.D., Tinker, L. ... & Seldin, M.F. (2012). Relationship between diabetes
risk and admixture in postmenopausal African-American and Hispanic-American women. Diabetologia 55: 1329-1337.
*Reiner, A.P., Ziv, E., Lind, D.L., Nievergelt, C.M., Schork, N.J., Cummings, S.R. ... & Kwok, P.Y. (2005). Population
structure, admixture, and aging-related phenotypes in African American adults: The Cardiovascular Health Study.
American Journal of Human Genetics 76: 463-477.
*Risch, N., Choudhry, S., Via, M., Basu, A., Sebro, R., Eng, C. ... & Burchard, E.G. (2009). Ancestry-related assortative
mating in Latino populations. Genome Biology 10(11): R132.
*Robbins, J.A., Qi, L., Garcia, L., Younger, J.W. & Seldin, M.F. (2015). Relationship of pain and ancestry in African
American women. European Journal of Pain 19: 601-610.
Roth, P.L., Bevier, C.A., Bobko, P., Switzer, F.S. & Tyler, P. (2001). Ethnic group differences in cognitive ability in
employment and educational settings: A meta‐analysis. Personnel Psychology 54: 297-330.
Ruiz-Linares, A., Adhikari, K., Acuña-Alonzo, V., Quinto-Sanchez, M., Jaramillo, C., Arias, W., Fuentes, M. et al. (2014).
Admixture in Latin America: Geographic structure, phenotypic diversity and self-perception of ancestry based on 7,342
individuals. PLoS Genetics 10(9): e1004572.

27
*Ruiz-Narváez, E.A., Bare, L., Arellano, A., Catanese, J. & Campos, H. (2010). West African and Amerindian ancestry
and risk of myocardial infarction and metabolic syndrome in the Central Valley population of Costa Rica. Human
Genetics 127: 629-638.
Russo, M.G., Di Fabio Rocca, F., Doldán, P., Gonzalo Cardozo, D., Dejean, C., Seldes, V. & Avena, S. (2016).
Evaluation of the minimum number of markers for individual ancestry estimation in an Argentinean population sample.
Revista del Museo de Antropología 9(1): 49-56.
*Salam, M.T., Avoundjian, T., Knight, W.M. & Gilliland, F.D. (2015). Genetic ancestry and asthma and rhinitis
occurrence in Hispanic children: Findings from the Southern California Children’s Health Study. PloS One 10(8):
e0135384.
Salzano, F.M. & Sans, M. (2014). Interethnic admixture and the evolution of Latin American populations. Genetics and
Molecular Biology 37: 151-170.
*Sánchez, E., Rasmussen, A., Riba, L., Acevedo‐Vasquez, E., Kelly, J.A., Langefeld, C.D. ... & Maradiaga‐Ceceña,
M.A. (2012). Impact of genetic ancestry and sociodemographic status on the clinical expression of systemic lupus
erythematosus in American Indian–European populations. Arthritis & Rheumatism 64: 3687-3694.
Scharf, J.M., Yu, D., Mathews, C.A., Neale, B.M., Stewart, S.E., Fagerness, J.A., Evans, P. et al. (2013). Genome-wide
association study of Tourette's syndrome. Molecular Psychiatry 18: 721-728.
*Schlesinger, D., Grinberg, L.T., Alba, J.G., Naslavsky, M.S., Licinio, L., Farfel, J.M., ... & Dos Santos, A.C.F. (2011).
African ancestry protects against Alzheimer's disease-related neuropathology. Epub 2011 Nov 8. [(Molecular Psychiatry
18: 79-85 (2013)].
Schmidt, F.L. & Hunter, J.E. (2014). Methods of Meta-Analysis: Correcting Error and Bias in Research Findings. Sage.
Shriner, D. (2013). Overview of admixture mapping. Current Protocols in Human Genetics: 1-23.
Shriver, M.D. & Kittles, R.A. (2004). Genetic ancestry and the search for personalized genetic histories. Nature Reviews
Genetics 5: 611-618.
*Sweeney, C., Wolff, R.K., Byers, T., Baumgartner, K.B., Giuliano, A.R., Herrick, J.S. ... & Slattery, M.L. (2007). Genetic
admixture among Hispanics and candidate gene polymorphisms: Potential for confounding in a breast cancer
study? Cancer Epidemiology Biomarkers & Prevention 16: 142-150.
*Tandon, A., Chen, C.J., Penman, A., Hancock, H., James, M., Husain, D. ... & Riche, D. (2015). African ancestry
analysis and admixture genetic mapping for proliferative diabetic retinopathy in African Americans. Investigative
Ophthalmology & Visual Science 56: 3999-4005.
Tate, B.W. & McDaniel, M.A. (2008, August). Race differences in personality: An evaluation of moderators and
publication bias. Paper presented at the Annual Meeting of the Academy of Management, Anaheim CA.
Telles, E. (2014). Pigmentocracies: Ethnicity, Race, and Color in Latin America. UNC Press Books.
Tucker-Drob, E.M. & Bates, T.C. (2015). Large cross-national differences in gene × socioeconomic status interaction on
intelligence. Psychological Science 27: 138-149.

28
*Valenzuela, C.Y. (1988). On sociogenetic clines. Ethology and Sociobiology 9: 259-268.
Van der Meer, T. & Tolsma, J. (2014). Ethnic diversity and its effects on social cohesion. Annual Review of Sociology
40: 459-478.
Van Ryn, M. (2002). Research on the provider contribution to race/ethnicity disparities in medical care. Medical Care
40(1): 1-140.
*Via, M., Gignoux, C.R., Roth, L.A., Fejerman, L., Galanter, J., Choudhry, S. ... & Martínez-Cruzado, J.C. (2011). History
shaped the geographic distribution of genomic admixture on the island of Puerto Rico. PLoS One 6(1): e16513.
Viechtbauer, W. (2015). metafor: Meta-Analysis Package for R (Version 1.9-8). Retrieved from https://cran.r-
project.org/web/packages/metafor/index.html
Winkler, C.A., Nelson, G.W. & Smith, M.W. (2010). Admixture mapping comes of age. Annual Review of Genomics and
Human Genetics 11: 65-89.
Wraw, C., Deary, I.J., Gale, C.R. & Der, G. (2015). Intelligence in youth and health at age 50. Intelligence 53: 23-32.
*Zou, J.Y., Park, D.S., Burchard, E.G., Torgerson, D.G., Pino-Yanes, M., Song, Y.S., Sankararaman, S., Halperin, E. &
Zaitlen, N. (2015). Genetic and socioeconomic study of mate choice in Latinos reveals novel assortment patterns.
Proceedings of the National Academy of Sciences 112: 13621-13626.