oa Within-household sampling: searching for a better methodconditional on household size information

Survey research has been used widely in various social sciences. A random selection of a survey respondent at the household level (so called within-household sampling) is critical for any valid statistical inference with the survey data. In this paper, we will review existing sampling methods. Some methods ensure the randomness, but require a lengthy and intrusive process and hence reduce cooperation. Some methods provide a quick and simple sampling at the expense of the randomness. Although household size information is collected in most of these methods, this information is not fully used. The question that these methods try to answer is how to randomly sample a person. In this paper, we argue that the method should be developed not to answer this question but to answer the question about how to randomly sample a person conditional on (or given) the household size. Compared to the current “one size fit all” methods, it is always better to use a method that allows for different ways of sampling for different household sizes. We then develop a sampling method that is conditional on the household size and use it for a survey in Qatar, a country with large household size population, a characteristic which is typical to the Middle East and Developing countries. We will show difficulties when current methods are applied to these countries and how our method can overcome these difficulties.