Title: Permutation test based on clustered data from a rotating sample plan
Authors: Jiahua Chen - University of British Columbia (Canada) [presenting]
Abstract: A classical problem in mathematical statistics is the hypothesis test. Given a data set, we wish to decide whether or not the distribution behind the data violates the model structure of interest. Such a simple task may demand complex solutions when a realistic yet comprehensive model is hard to find. In an applied project, we have observations on samples from several connected populations. Due to a rotating sampling plan,random effects are suggested in its longitudinal direction as well as in cross-sectional respect. Besides, strong parametric model assumptions should be discouraged. It is difficult to quantify or model these random effects. The asymptotic theory becomes hard to develop, and therefore a good approximation hard to find for the distribution of the test statistics. We develop a permutation scheme to the symmetric in the data structure. The resulting test, therefore, automatically has the right size. Combined with a semi-parametric density ratio model and the composite likelihood approach, the proposed tests are found to work well for the targeted applications.