CFE 2019: Start Registration
View Submission - CMStatistics
Title: Optimal sample allocation for two-phase designs in cluster correlated data settings Authors:  Claudia Rivera-Rodriguez - The University of Auckland (New Zealand) [presenting]
Sebastien Haneuse - Harvard TH Chan School of Public Health (United States)
Abstract: Large amount of the research in survey sampling has been directed towards improving estimates using information available for the entire population. The efficiency of such methods depends on several factors such as the information available for the entire population, the sampling strategy, the sample size, etc. There is no an absolutely optimal design, but under certain principles and restrictions, a well designed sampling strategy can be implemented. In two-phase designs, efficiency is gained with stratification by auxiliary information known early in the design. There is a number of reasons why investigators may want to stratify: it offers gains in efficiency when the target variable behaves differently between strata and estimates can be obtained for each strata. A further way to gain efficiency is by optimally allocating the resources. For example, conditional on a given sample size or a given precision, what is the optimal allocation of sample sizes? Large amount of the research on optimal allocation has been directed towards estimation of totals or functions of totals. However, in many instances inference is concerned with regression parameters from data that arises from a correlated setting. We examine the impact of ignoring the correlation in when allocating thee resources. Using theory from sampling survey, we extend and propose different allocation methods that allow for correlated data.