Latent Variable Models in the Investigation of Salary Discrimination: Theory and Practice
讨论了在工资歧视诉讼中，当重要能力变量未测量或测量有误差时，使用潜变量路径模型替代传统回归分析的理论与实践，并用真实数据演示了模型应用。
Statistical methods play an important role in salary discrimination litigation. Regression analysis is the most widely used method in this application. In regression analysis, group differences in salary are evaluated among employees who have been matched on measured “merit” variables (e.g., years of job experience). Difficulties often arise because important merit variables may be unmeasured, or because the available measures of merit are fallible. An alternative approach in this case lies in the use of several latent variable path models that have been proposed for the salary problem (Birnbaum, 1979; McFatter, 1987). The theoretical and practical implications of these models are discussed. The use of the models is illustrated in real data.