Following earlier work [19], predicted probabilities of separation were derived from logistic regression models predicting parental separation from parental alcoholism and other family background variables presumed to predate separation. To reduce bias associated with missing data from fathers [16], we limited predictors to those available by mother report, specifically, mother self-report AD symptoms and mother report of coparent alcohol problems. Additional predictors included maternal age at twins’ birth, paternal education, maternal smoking, and two-way interactions between parental alcohol problems and (i) maternal age at twins’ birth and (ii) paternal education. Note that paternal rather than maternal education was modeled for purposes of PSA as paternal education was more strongly predictive of separation. Maternal history of smoking was likewise modeled rather than either parent history of smoking.
