ESTIMATING MEASUREMENT ERROR USING MULTIPLE INDICATORS AND SEVERAL POINTS IN TIME.

The article discusses the estimation of measurement error using multiple indicators. A number of recent papers have considered various approaches to measurement error by utilizing explicit causal models linking unmeasured variables to their measured indicators. In general, we know that unless there are a large number of measured variables relative to unmeasured variables it will be difficult, if not impossible, to reach definite conclusions, since the existence of unmeasured variables in a causal system introduces a relatively large number of unknowns, thereby necessitating additional assumptions which are often rather implausible. In this article, the author would like to combine selected features of the arguments developed by researchers Herbert L. Costner and David R. Heise, while emphasizing the flexibility of the general approach that is common to both papers. When we study the implications of non- random measurement errors, even where we have multiple indicators and several time periods, we begin to see the importance of careful measurement at the data collection stage, as well as working with large samples that have been carefully designed.