To own Japanese–White pairing habits, i next reestimated the coefficients and you may chance rates towards sample excluding combat-fiance intermarriages
The term of interest is ? rR , from which the odds ratios were calculated. The constraints used in the two models for this effect are described below as design matrices. In the matrix, intermarriages between non-White groups were constrained to be equal (?step step one0). Models 1 and 2 differ by the specification of the interaction of husbands’ and wives’ race but include the same dimensions selected on the basis of findings in preliminary analyses. The simplified version of the structure of the models (Equation 1), with NrReEvG as the cell frequency, ?o as the intercept, and abbreviated indices and summation signs for categories in each dimension, is as follows:
Model 1 tests the significance of the influences of composition and educational pairing. A BIC value of less than 0 in Model 1 would indicate that the influences are statistically significant. This is because preliminary analyses showed that a simple model that includes only husbands’ race (? r ) and wives’ race (? R ) and their interaction term (? rR ), which excludes other measures from the model, yielded BIC values much greater than zero, which indicate a fit worse than the saturated model (with all multiway interaction effects). Model 2 tests the presence of gender asymmetry in the racial pairing patterns of the resettlement cohort. Intererican wives and White husbands might have become more common than intererican husbands and White wives after WWII. The War Bride Act could have legitimized the former type of intermarriage, and, in addition, had brought in intermarriages of Japanese war-brides and White veteran husbands from overseas.