2016/03/01 by Lubna Amro, Markus Pauly, Amro, Lubna +1
Computer Science · Mathematics · #62G09 #62G10 #Bayesian Methods and Mixture Models #FOS: Mathematics #Statistical Methods and Inference #Statistical Methods in Clinical Trials #Statistics Theory (math.ST)
paper · pdf · doi:10.48550/arxiv.1603.00214
openalex publication_date 2016/03/01 · openalex created_date 2022/10/05 · openalex updated_date 2026/08/01
Various statistical tests have been developed for testing the equality of\nmeans in matched pairs with missing values. However, most existing methods are\ncommonly based on certain distributional assumptions such as normality,\n0-symmetry or homoscedasticity of the data. The aim of this paper is to develop\na statistical test that is robust against deviations from such assumptions and\nalso leads to valid inference in case of heteroscedasticity or skewed\ndistributions. This is achieved by applying a novel randomization approach. The\nresulting test procedure is not only shown to be asymptotically correct but is\nalso finitely exact if the distribution of the data is invariant with respect\nto the considered randomization group. Its small sample performance is further\nstudied in an extensive simulation study and compared to existing methods.\nFinally, an illustrative data example is analyzed.\n