Research Note: Reliability and Dimensionality of Stimulus Ratings
传统信度统计方法将信度等同于一致性，但有些一致性并不理想；同样，对广告、策略和人的属性进行因子分析时，得到的因子更多描述评分者而非被评对象，可能导致错误且代价高昂的决策。
The traditional statistics for estimating reliability have misled most of us who use stimulus ratings into thinking our data are more reliable than they really are. The problem is that the standard statistics equate reliability with consistency, even though there are kinds of consistency which none of us would consider desirable. Similarly, conventional research designs for factor analyzing the attributes of ads, strategies, and people produce factors more descriptive of the raters than of the objects rated. These customary ways of doing things may lead to decisions that are not only wrong but confusing and potentially very costly.