论文部分内容阅读
最小化法是一种动态随机化算法,它可以在较小的样本量下保证各处理组间和控制因素水平下的样本均衡.介绍了最小化算法的原理,并利用形式化方法对其进行了详细的描述,对算法中较为核心的均衡度量和不均衡计算方法做了说明.提出最小化法存在的三个缺陷:不可重复、不支持非平衡设计和计算复杂度高.针对上述缺陷,通过加入基于种子的随机序列来控制分组概率,从而解决不可重复的问题,通过用处理组间样本比例来修正优先分配概率解决非平衡设计,通过设计一套计算机函数来解决复杂度,最终使这些问题都得到很好地解决.“,”Minimization can meet the requirements of keeping balance between interventions and control factors when a study has small quantities of samples.For this reason,this article explained the theory of minimization and described the details by formal method.However,there are three weaknesses of this algorithm:the process is unrepeatable,unbalance design is unsupportable,and the calculation is complex.In order to resolve these problems,the author gave some advice that using a fixed random number list calculated by a certain seed to control the allocation probability,timing priority probabilities and sample ratios of each intervention groups to adjust priority probabilities,and developing a computer function to reduce the difficulties of algorithm application.