Assessing and relaxing assumptions in quasi-simplex models

Abstract

The quasi-simplex model (QSM) makes use of at least three repeated measures of the same variable to estimate reliability. The model has rather strict assumptions and ignoring them may bias estimates of reliability. While some previous studies have outlined how several of its assumptions can be relaxed, they have not been exhaustive and systematic. Thus, it is unclear what all the assumptions are and how to test and free them in practice. This chapter will addresses this situation by presenting the main assumptions of the quasi-simplex model and the ways in which users can relax these with relative ease when more than three waves are available. Additionally, by using data from the British Household Panel Survey we show how this is practically done and highlight the potential biases found when ignoring the violations of the assumptions. We conclude that relaxing the assumptions should be implemented routinely when more than three waves of data are available.