Dealing with incomplete data in questionnaires of food and alcohol consumption

Missing data feature in most large-scale surveys of human populations, especially when extensive questionnaires are administered. We describe the established single-imputation procedures and explore multiple-imputation procedures for the missing values in the blocks of questions related to alcohol consumption in the UK Women’s Cohort Study. We demonstrate how multiple imputation can be developed by adapting the established single-imputation procedures, ridding them of some of their weaknesses.