Specifically, the weights for the participants were calculated as the reciprocal of the
probability of completing the study, with the probabilities being estimated from a logistic regression with completion or non-completion as outcome, age, baseline vegetable con-sumption, marital and smoking status being independent variables.