Raking Method as a Tool for Improving Representativeness in Non-Probability Studies

This is a methodological review focused on raking, or iterative proportional fitting, as a tool for improving representativeness in studies with non-probability sampling. The paper synthesizes the theoretical foundations, practical considerations, and applications of raking in biomedical research. The method operates by iteratively adjusting sample weights so that the marginal distributions of selected variables match the known distributions of the target population. Its implementation requires reliable auxiliary information about the population of interest and careful selection of adjustment variables. The review addresses critical aspects such as weight quality evaluation, management of extreme values, and computational considerations in raking implementation. The method's advantages are discussed, including its capacity to simultaneously adjust multiple variables and its applicability when only marginal information about the population is available. Its limitations are also examined, such as the potential generation of extreme weights and dependence on precise population data. Finally, practical examples are presented in various contexts, from hospital studies to research in university populations, demonstrating the method's versatility. The application of raking has proven particularly valuable in epidemiological and health services studies, where non-probability samples are common. This review provides a comprehensive methodological guide for researchers seeking to implement raking, emphasizing the importance of rigorous application and transparent documentation.