CONSTRUCTING EXACT SIGNIFICANCE TESTS WITH RESTRICTED RANDOMIZATION RULES
成果类型：
Article
署名作者：
MEHTA, CR; PATEL, NR; WEI, LJ
署名单位：
Indian Institute of Management (IIM System); Indian Institute of Management Ahmedabad; University of Michigan System; University of Michigan
刊物名称：
BIOMETRIKA
ISSN/ISSBN：
0006-3444
发表日期：
1988
页码：
295302
关键词：
摘要：
In a clinical trial comparing two treatments, suppose that subjects are assigned sequentially to the treatment groups by a restricted randomization rule. Under the randomization model, an efficient recursive algorithm is provided to generate the exact permutational distribution for linear rank statistics, given the final imbalance between the number of patients in the two groups. If only the significance level of the test is required, the efficiency of the algorithm can be improved significantly. In general, by the time the size of the trial reaches 30, the large-sample approximation to the permutation distribution (Wei, Smyth and Smith, 1986) is satisfactory. The new procedure is illustrated by censored survival data from a prostatic cancer clinical trial. We show that, if the study subjects are not a random sample from a well-defined population, then the treatment allocation rule used at the design stage should not be ignored in the analysis.