Repeated Observations In The Small-scale Clinical Trials Evaluating The Therapeutic Efficacy Of Statistical Methods

This paper aims at the statistical methods for analysing repeated measurements data from phase I and II clinical trials (first period and the first phase of the second period according to new Chinese regulations) . A statistical test for testing efficacy of treatments is given for data from clinical trials where there is only one or two patients of rare diseases. A mixed homogeneous finite Markov chain model is brought forward for dealing with situations where there are two groups of patients. And a statistical test based on this model is also given. In this method, the assumptions concerning the data of normality, independence and homoscedasticity are moved out. Incomplete data, different sample sizes and weighted observations at different time points in accordance with concrete background are all accommodated. Monte Carlo simulations show agreeably to theoretical analysis that this method yields fairly stable test levels and has higher power than its classical competitor split-plot ANOVA.