Table 1.
Results from models 1–5 for the top five traits that showed significant evidence of gene–environment correlations that extended beyond the family level across geographic regions
| 1. Individual-level polygenic score only | 2. Within + between family | 3. Within + between region | 4. Within + between family + between region | 5. Within family fixed effect split into within and between region fixed effect | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Estimate (SE) | t-value (p) | Estimate (SE) | t-value (p) | Estimate (SE) | t-value (p) | Estimate (SE) | t-value (p) | Estimate (SE) | t-value (p) | |
| BMI | ||||||||||
| Within/individual level | −0.10 (0.005) | −19.29 (2 × 10−82) | −0.06 (0.008) | −7.77 (8 × 10−15) | −0.07 (0.005) | −14.23 (7 × 10−46) | −0.05 (0.008) | −6.04 (1 × 10−9) | −0.04 (0.009) | −4.21 (3 × 10−5) |
| Between family | – | – | −0.11 (0.006) | −18.22 (1 × 10−73) | – | – | −0.08 (0.006) | −13.26 (6 × 10−40) | – | – |
| Between region | – | – | – | – | −0.20 (0.007) | −29.75 (1 × 10−181) | −0.16 (0.011) | −14.10 (6 × 10−45) | −0.34 (0.060) | −5.73 (1 × 10−8) |
| Marginal R2 (fixed effects only) | 0.016 | 0.018 | 0.033 | 0.033 | 0.009 | |||||
| Conditional R2 (fixed and random effects) | 0.291 | 0.292 | 0.300 | 0.300 | 0.302 | |||||
| Waist circumference | ||||||||||
| Within/individual level | −0.07 (0.004) | −16.55 (3 × 10−61) | −0.05 (0.008) | −7.02 (2 × 10−12) | −0.05 (0.004) | −12.06 (2 × 10−33) | −0.04 (0.008) | −5.61 (2 × 10−8) | −0.04 (0.009) | −4.10 (4 × 10−5) |
| Between family | – | – | −0.08 (0.005) | −15.43 (2 × 10−53) | – | – | −0.06 (0.005) | −1.95 (8 × 10−28) | – | – |
| Between region | – | – | – | – | −0.15 (0.006) | −24.56 (3 × 10−127) | −0.12 (.010) | −11.25 (3 × 10−29) | −0.26 (0.055) | −4.70 (3 × 10−6) |
| Marginal R2 (fixed effects only) | 0.217 | 0.217 | 0.227 | 0.227 | 0.216 | |||||
| Conditional R2 (fixed and random effects) | 0.411 | 0.411 | 0.424 | 0.424 | 0.428 | |||||
| Household income | ||||||||||
| Within/individual level | 0.12 (0.005) | 25.10 (8 × 10−138) | 0.06 (0.01) | 7.12 (1 × 10−12) | 0.07 (0.005) | 15.25 (2 × 10−52) | 0.04 (0.01) | 4.46 (8 × 10−6) | 0.03 (0.01) | 3.30 (1 × 10−3) |
| Between family | – | – | 0.15 (0.01) | 25.47 (6 × 10−141) | – | – | 0.09 (0.01) | 15.35 (6 × 10−53) | – | – |
| Between region | – | – | – | – | 0.33 (0.01) | 44.39 (<10−310) | 0.31 (0.01) | 25.69 (3 × 10−143) | 0.41 (0.09) | 4.58 (5 × 10−6) |
| Marginal R2 (fixed effects only) | 0.127 | 0.130 | 0.196 | 0.197 | 0.127 | |||||
| Conditional R2 (fixed and random effects) | 0.320 | 0.321 | 0.367 | 0.368 | 0.377 | |||||
| Time spent watching television | ||||||||||
| Within/individual level | −0.14 (0.005) | −27.74 (1 × 10−167) | −0.06 (0.008) | −6.60 (4 × 10−11) | −0.10 (0.005) | −2.74 (5 × 10−95) | −0.04 (0.009) | −4.51 (6 × 10−6) | −0.03 (0.009) | −3.32 (9 × 10−4) |
| Between family | – | – | −0.17 (0.006) | −29.07 (3 × 10−182) | – | – | −0.13 (0.006) | −21.95 (1 × 10−105) | – | – |
| Between region | – | – | – | – | −0.29 (0.007) | −42.51 (<10−310) | −0.28 (0.011) | −24.24 (4 × 10−128) | −0.33 (0.070) | −4.62 (4 × 10−6) |
| Marginal R2 (fixed effects only) | 0.044 | 0.050 | 0.080 | 0.082 | 0.029 | |||||
| Conditional R2 (fixed and random effects) | 0.267 | 0.273 | 0.279 | 0.284 | 0.299 | |||||
| Whole-body fat mass | ||||||||||
| Within/individual level | −0.08 (0.005) | −16.20 (8 × 10−59) | −0.05 (0.008) | −6.32 (3 × 10−10) | −0.06 (0.005) | −11.85 (3 × 10−32) | −0.04 (0.008) | −5.02 (5 × 10−7) | −0.03 (0.009) | −3.62 (3 × 10−4) |
| Between family | – | – | −0.09 (0.006) | −15.44 (2 × 10−53) | – | – | −0.07 (0.006) | −11.06 (2 × 10−28) | – | – |
| Between region | – | – | – | – | −0.17 (0.007) | −25.14 (8 × 10−133) | −0.13 (0.011) | −12.02 (3 × 10−33) | −0.26 (0.058) | −4.60 (5 × 10−6) |
| Marginal R2 (fixed effects only) | 0.070 | 0.071 | 0.081 | 0.082 | 0.063 | |||||
| Conditional R2 (fixed and random effects) | 0.328 | 0.327 | 0.339 | 0.338 | 0.338 | |||||
This table contains results from linear mixed effect models, and reports the fixed effect estimates, their standard errors (SE), and the t-value and corresponding uncorrected one-sided P-value of the significance of the fixed effect.