In most surveys the data are modified in some way during processing. For example, inconsistencies in the data structure will have been cleaned up and new variables will have been derived from combinations of responses to a set of questions. For example, a variable which describes the family type may have been constructed based on the response to a single question asking about family type or it may have derived from a series of questions about each individual in the household and how they are related to each other.

Other processes that may have been carried out to transform the data will be to impute for missing data. This is sometimes built into the design of the survey such as in education assessments using direct testing methods which routinely use a methodology where scores are imputed based on the responses of others with similar characteristics. Have the same rules been applied in all the countries?