Regression results for women and men are shown in Table 2a,b, respectively. We first estimated the association of demographic covariates with PYDD. Higher PYDD was associated with not being married in both sexes, higher income in women and lower education in men. We thus used the fit of a model consisting of marital status, income, and education as the baseline against which subsequent models were compared (Models F1 and M1). When simultaneously adjusting for the aforementioned demographic covariates, age at interview was not a significant predictor of PYDD and thus was not included in subsequent models.
