Potter R H, McDonald R E
J Dent Educ. 1985 Mar;49(3):145-53.
The notion of prediction implies causation. Path and structural models explore the causal links rather than mere empirical relationships between variables. This technique involves a breakdown of correlations; it differs from correlation and regression methods in that it provides relevant information in the presence of important but unobserved (latent) explanatory variables and of measurement errors in the data. It also allows for more than one regression analysis simultaneously and affords inference through tests of the model. In this study, latent abilities of dental students were analyzed as causes and professional achievements as effects, with preadmission performances as indicators of latent abilities. A model with three constructs is consistent with the observed data. The results demonstrate that whereas correlation analysis presents limitations in interpretation, structural analysis focuses clearly on the direct impact of the quality of dental school education, rather than preadmission background, on clinical performance and board success as measures of future professional performance.
预测的概念意味着因果关系。路径模型和结构模型探索的是因果联系，而非变量之间单纯的经验关系。这项技术涉及对相关性的分解；它与相关分析和回归方法的不同之处在于，在存在重要但未观测到的（潜在）解释变量以及数据存在测量误差的情况下，它能提供相关信息。它还允许同时进行多个回归分析，并通过模型检验进行推断。在本研究中，将牙科学生的潜在能力作为原因进行分析，将专业成就作为结果进行分析，入学前成绩作为潜在能力的指标。一个具有三个结构的模型与观测数据相符。结果表明，虽然相关分析在解释方面存在局限性，但结构分析明确聚焦于牙科学院教育质量而非入学前背景对临床成绩和委员会考试成功（作为未来专业表现的衡量指标）的直接影响。
