Analysis of Experimental Data with Repeated Measurements

摘要

Experimental data often consist of serial measurements on subjects after a treatment. Typical questions concerning such data are: (A) Do subjects really react to treatment or are the fluctuations just random? (B) What are the numerical characteristics of the response? (C) Is the response identical in all groups? Differences between the individuals in the dynamics of the reaction make it difficult to apply standard statistical procedures. This paper proposes to answer questions (A) and (B) at the individual level, then to give an answer to (C) on the basis of this information. This kind of analysis may be useful since it can separate subjects giving response from those that do not and can identify individual response patterns and compare treatments with respect to each numerical characteristic separately. To answer question (A), a permutation test is proposed and its power is evaluated by simulation.