Using discrete time survival analysis (hazard models) [33], the hazard of attaining a milestone (the risk of attainment among those who had not yet reached the milestone) was estimated. Survival analyses handle right censoring due to failure to reach a milestone because of attrition or study end. Missing data were modeled for individuals who already initiated a given event at study outset. Group differences in hazard functions were tested for sex, onset age, and race (controlling for age and other predictors). To depict time to milestone attainment, age of attainment was predicted from sex, from age of onset, and from race using a series of general linear models (GLMs), and age of attainment was plotted for each milestone by sex, age of onset, and race. Although this method does not account for censoring, it is intuitively appealing and these figures are more transparent than hazard function graphs.
