Some contributions on the use of auxiliary information in sample surveys.
Mulati, Omukoba Nyukuri
MetadataShow full item record
In this project, we assume that we have a population from which a sample is taken and that not all sample units are observed... It is also assumed that auxiliary information is available for all units in the population. In this case to make inference about the population, we encounter two procedural problems, of estimation and imputation. The big question is do we use auxiliary information for imputation or do we use it for estimation? We have tackle this problem by considering the following strategies: use of hot deck imputation technique to fill in missing values and then of ratio imputation method to impute the missing values and then use of an expansion estimator to estimate the population total. We have suggested some recommendation on when to use auxiliary information. In particular, when non-response is random the use of auxiliary information at imputation is preferred. This is evident from both our theoretical and empirical study carried out to support our decisions and further recommend others.