ABSTRACT
This chapter presents several papers that closely related models for the estimation of reliability coefficients and the measurement-error variances from univariate panel data collected at three or more time points. It examines an instance in which the four-wave model fits the data very well; yet, when it is elaborated with a tau-equivalent measure at one of the time points, the fit is very poor. The chapter discusses possible violations of the assumptions of the model, proposes an alternative model that resolves the apparent difficulty, and explains that some types of correlated disturbances cannot be detected by the goodness-of-fit test for the basic four-wave model. In a study of the properties of several measures of scientific productivity, the chapter explores the reliability and error variance of article counts taken from the Science Citation Index as a measure of the true article production of individual scientists.
