Mikeal R L
Am J Hosp Pharm. 1980 Apr;37(4):541-8.
The conceptual foundations for some of the most widely used research designs are presented, and the various design models are analyzed. Research designs should specify the stimulus and response, the time-order sequence, the sampling and allocation procedures, and the generalized method of analysis. Results of symbolization, seriation, and classification investigations cannot be generalized beyond the study units investigated. Cohort or case-control investigations prevent determination of causality. Cause can be logically proven if the design uses random assignment of study units to study groups and of study groups to stimulus conditions. Crossover designs offer no advantages over similarly constructed noncrossover designs, and introduce new sources of invalidity.
介绍了一些最广泛使用的研究设计的概念基础，并分析了各种设计模型。研究设计应明确刺激和反应、时间顺序序列、抽样和分配程序以及通用的分析方法。符号化、序列化和分类调查的结果不能推广到所研究的单位之外。队列或病例对照调查无法确定因果关系。如果设计使用将研究单位随机分配到研究组以及将研究组随机分配到刺激条件，则可以从逻辑上证明因果关系。交叉设计与类似构建的非交叉设计相比没有优势，并且会引入新的无效来源。
