Table 2.
Regression analyses assessing the extent to which schooling attenuates and moderates the associations of early-life stature and child IQ with adult IQ
|
Brazil |
Guatemala |
The Philippines |
South Africa |
|||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Female (n=249) | Male (n=201) | Female (n=229) | Male (n=196) | Female (n=590) | Male (n=686) | Female (n=247) | Male (n=216) | |||
| Hypothesis: controlling for schooling will attenuate the strength of the association between early-life stature and adult cognition but not childhood cognition and adult cognition | ||||||||||
| Adult IQ | ||||||||||
| Path J | ||||||||||
| Schooling | 2·86 (2·08 to 3·63) | 3·75 (3·03 to 4·47) | 0·74 (0·58 to 0·89) | 0·83 (0·64 to 1·03) | 2·40 (2·08 to 2·72) | 1·85 (1·60 to 2·11) | 2·75 (1·79 to 3·71) | 1·69 (0·89 to 2·48) | ||
| Paths H and J | ||||||||||
| Early-life stature | 1·35 (0·02 to 2·68) | −0·29 (−1·90 to 1·31) | −1·35 (−2·90 to 0·21) | 1·88 (0·22 to 3·53) | 1·67 (0·72 to 2·62) | 0·19 (−0·72 to 1·10) | 0·05 (−1·82 to 1·91) | 1·12 (−0·49 to 2·72) | ||
| Schooling | 3·70 (2·99 to 4·42) | 2·89 (2·08 to 3·70) | 2·06 (1·64 to 2·48) | 2·12 (1·59 to 2·65) | 3·06 (2·63 to 3·49); p<0 | 2·45 (2·11 to 2·79) | 4·16 (2·72 to 5·66) | 2·55 (1·34 to 3·76) | ||
| Paths I and J | ||||||||||
| Child IQ | 7·31 (5·57 to 9·22) | 3·90 (2·18 to 5·62) | 4·88 (3·38 to 1·25) | 3·74 (1·87 to 5·62) | 4·96 (3·94 to 5·99) | 5·77 (4·74 to 6·66) | 2·77 (0·88 to 4·54) | 3·95 (2·28 to 5·62) | ||
| Schooling | 2·63 (1·92 to 3·33) | 2·48 (1·66 to 3·30) | 1·59 (1·19 to 1·99) | 1·82 (1·27 to 2·37) | 2·46 (2·03 to 2·89) | 1·77 (1·45 to 2·10) | 3·63 (2·14 to 5·13) | 1·99 (0·81 to 3·17) | ||
| Paths H, J, and I | ||||||||||
| Early-life stature | 0·41 (−0·71 to 1·54) | −0·14 (−1·67 to 1·38) | −1·86 (−3·29 to −0·44) | 1·21 (−0·45 to 2·86) | 0·78 (−0·14 to 1·69) | −0·35 (−1·18 to 0·47) | 0·01 (−1·82 to 1·85) | 0·84 (−0·69 to 2·38) | ||
| Child IQ | 7·31 (5·41 to 9·06) | 3·90 (2·18 to 5·62) | 5·13 (3·63 to 6·63) | 3·46 (1·44 to 5·33) | 4·82 (3·65 to 5·84) | 5·77 (4·88 to 6·66) | 2·77 (0·88 to 4·54) | 3·80 (2·13 to 5·47) | ||
| Schooling | 2·63 (1·93 to 3·34) | 2·50 (1·66 to 3·34) | 1·68 (1·28 to 2·08) | 1·78 (1·23 to 2·33) | 2·43 (2·01 to 2·86) | 1·78 (1·46 to 2·11) | 3·63 (2·14 to 5·13) | 1·98 (0·80 to 3·16) | ||
| Hypothesis: the associations of early-life stature and of child IQ with adult IQ will be moderated by schooling | ||||||||||
| Adult IQ | ||||||||||
| Paths H, J, and H*J | ||||||||||
| Early-life stature | 1·96 (−1·93 to 5·85) | 4·38 (−1·48 to 10·23) | −0·56 (−2·70 to 1·58) | 0·88 (−2·08 to 3·84) | −0·15 (−3·70 to 3·40) | 0·47 (−2·32 to 3·26) | 3·24 (−20·23 to 26·71) | 3·61 (−11·52 to 18·75) | ||
| Schooling | 3·68 (2·98 to 4·37) | 2·53 (1·56 to 3·50) | 1·47 (0·27 to 2·66) | 2·63 (1·26 to 4·01) | 3·50 (2·57 to 4·43) | 2·38 (1·60 to 3·15) | 3·89 (1·27 to 6·52) | 2·14 (−0·58 to 4·87) | ||
| Early-life stature*schooling (interaction) | −0·06 (−0·49 to 0·36) | −0·44 (−1·06 to 0·17) | −0·22 (−0·63 to 0·19) | 0·19 (−0·28 to 0·65) | 0·18 (−0·15 to 0·51) | −0·03 (−0·29 to 0·24) | −0·27 (−2·24 to 1·70) | −0·22 (−1·52 to 1·09) | ||
| Paths I, J, and I*J | ||||||||||
| Child IQ | 2·70 (−3·02 to 8·43) | −2·96 (−11·08 to 5·15) | 2·25 (0·00 to 4·50) | 1·02 (−2·30 to 4·32) | 0·00 (−3·50 to 3·36) | 5·33 (2·81 to 7·99) | 7·18 (−6·30 to 20·66) | 14·29 (0·30 to 28·42) | ||
| Schooling | −0·25 (−4·02 to 3·52) | −1·42 (−6·41 to 3·57) | −3·44 (−6·56 to −0·32) | −1·65 (−5·11 to 1·81) | −0·66 (−2·75 to 1·42) | 1·54 (−0·06 to 3·15) | 6·55 (−2·39 to 15·50) | 7·86 (−0·07 to 15·78) | ||
| Child IQ*schooling (interaction) | 0·48 (−0·16 to 1·11) | 0·62 (−0·16 to 1·40) | 0·63 (0·25 to 1·00) | 0·43 (0·00 to 0·86) | 0·44 (0·15 to 0·73) | 0·00 (−0·15 to 0·30) | −0·38 (−1·39 to 0·76) | −0·91 (−2·03 to 0·30) | ||
| Paths H, I, I*J, and H*J | ||||||||||
| Early-life stature | 1·46 (−2·03 to 4·96) | 7·02 (1·01 to 13·03); p<0·05 | −1·5 (−3·50 to 0·49) | 1·0 (−2·06 to 4·06) | 0·47 (−3·08 to 4·02) | −0·02 (−2·70 to 2·67) | 3·31 (−19·87 to 26·48) | 1·11 (−13·47 to 15·69) | ||
| Schooling | −0·81 (−4·69 to 3·07) | −2·16 (−7·01 to 2·69) | −2·79 (−6·16 to 0·57) | −1·54 (−5·88 to 2·79) | −0·53 (−3·17 to 2·11) | 1·38 (−0·67 to 3·43) | 6·32 (−2·80 to 15·45) | 7·38 (−1·29 to 16·05) | ||
| Child IQ | 1·75 (−4·13 to 7·47) | −2·96 (−10·61 to 4·52) | 2·75 (0·50 to 5·00) | 0·60 (−3·02 to 4·38) | −0·15 (−3·80 to 3·65) | 5·33 (2·52 to 7·99) | 7·31 (−6·30 to 20·92) | 13·53 (−0·76 to 27·82) | ||
| Early-life stature*schooling (interaction) | −0·09 (−0·47 to 0·28) | −0·68 (−1·31 to −0·06) | 0·00 (−0·38 to 0·38) | 0·06 (−0·42 to 0·54) | 0·03 (−0·30 to 0·36) | −0·03 (−0·29 to 0·22) | −0·28 (−2·22 to 1·67) | −0·03 (−1·29 to 1·23) | ||
| Child IQ*schooling (interaction) | 0·64 (−0·00 to 0·64) | 0·62 (−0·16 to 1·40) | 0·50 (0·13 to 0·88) | 0·45 (−0·00 to 1·06) | 0·44 (0·15 to 0·73) | 0·00 (−0·15 to 0·30) | −0·38 (−1·51 to 0·76) | −0·76 (−1·98 to 0·30) | ||
Data are β (95% CI). Paths are detailed in figure 1. Early-life stature was measured at age 2 years except in Pelotas (at age 1 year). Child IQ was measured at age 3–7 years (Guatemala), 4 years (Brazil and South Africa), and 8·5 years (the Philippines). Estimates are years of schooling per height-for-age Z score or per site-specific and sex-specific SD of child IQ. Analyses were adjusted for maternal schooling attainment (years), height (cm), wealth (quintile, categorical), and birth order (first, second, third, or fourth or later). Results for Guatemala were also adjusted for birth year and treatment assignment. Results for Brazil were weighted to reflect sampling.