Using structural equation models to evaluate the magnitude of measurement error in blood pressure.

This article aims to compare alternative methods for estimating the quality of blood pressure measurements. Traditional within-subject variance estimates in mixed analysis of variance models are compared to multiple-group multitrait-multimethod models, which are a particular case of mean-and-covariance-structure confirmatory factor analysis models. Confirmatory factor analysis models belong to the family of structural equation models and were specifically developed to analyse psychosociological traits measured by tests or surveys, but they have also proved suitable for evaluating the quality of blood pressure measurements. Confirmatory factor analysis models are less restrictive and provide more detailed information than traditional approaches, enable researchers to compute weighted averages of individual measures with optimal measurement quality, make it easier to correct the biasing effects of measurement error on the results of substantive studies, and make the use of equivalent replicated measures unnecessary under certain conditions.